The III-nitride semiconductor system has been and will most probably continue to be the dominant material in optoelectronics technology. As a classical representative of the third-generation of semiconductor materials, GaN-based optoelectronic materials have already exhibited their importance in applications including green/blue/visible/UV high-brightness light-emitting diodes (LEDs) and laser diodes (LDs) in solid-state-lighting devices with higher quantum efficiency [1-3]. InGaN/GaN multiple quantum wells (MQWs) are a crucial issue of the active region for optoelectronic devices which can be tuned according to indium components in the InGaN alloy system [4, 5], for example, blue light is emitted from the active region when the indium component is between 0.15 to 0.20 in In_XGa_1-XN/GaN MQWs LED. As is known, the best devices come from the highest quality epitaxial device layers. However, there is still extensive discussion on what mechanisms lead to the changing internal quantum efficiency (IQE) with different wavelengths and operating conditions. For example, efficiency-droop effect in blue LED under high-current injection, light extraction efficiency and lower external quantum efficiency (EQE). Owing to difficulty in preparation of bulk materials, nearly all epifilms are heteroepitaxially grown on foreign substrates, such as sapphire, SiC, and Si substrates, which results in threading dislocations and residual strains occurring greatly in active region of LEDs. The most important issue hampering the applications and advances in solid-state lighting to LED is dislocation and residual strains existing in epifilms, which are attributed to larger parameter mismatch and stronger polarization between epifilms and the sapphire substrate. Numerous techniques and investigations have been put to use for obtaining higher crystalline quality and EQE, including patterned sapphire substrate (PSS) and nano-micrometer hybrid patterned substrates [6, 7], whereas they have rarely shown the oscillatory structure of photoluminescence at room temperature. Nowadays PSS is de facto a standard for growth of III-Nitride heterostructures for high brightness LEDs because of improving the strain relaxation of epifilms and light extraction efficiency. However, up to now there are few studies concerning the crystalline quality, surface morphology, luminescence properties and residual strain of blue-LED grown on patterned sapphire substrate. In this paper, we have investigated optical properties of blue light LEDs using InGaN/GaN (indium component is 17.43%) MQWs as active region grown on PSS by low-pressure metal-organic chemical vapor deposition (MOCVD), and studied further the influence on the crystal quality, the threading dislocation density, surface morphology, residual strain existing in the active region and the optical properties of GaN epilayer on c-plane patterned sapphire substrates.

As shown in Fig. 1, the conventional and patterned sapphire substrate LED are designed to be grown on a c-plane sapphire substrate using a cold-wall showerhead low pressure metal-organic chemical vapor deposit (LPMOCVD) system. The pressure of the LPMOCVD growth system was about 5.2 × 10⁻³ Torr. Hydrogen was used as carrier gas, triethylgallium (TEGa), triethylindium (TEIn) and ammonia (NH₃) were used as Ga, In and N sources, respectively. A thin low-temperature AlN nucleation layer and a high-temperature AlN nucleation layer were grown on sapphire substrates, as shown in Fig. 1. Followed by a 2-μm-thick undoped GaN layer, and a 2.5-μm-thick n-type GaN (n doping = 1.0 × 10¹⁸/cm³). The active region of the conventional and PSS LEDs include five 2.5-nm-thick In_0.17Ga_0.83N QWs separated by six 10-nm-thick GaN barriers. A 10-nm-thick p-type Al_0.15Ga_0.85N electron blocking layer (EBL) (p doping = 8.0 × 10¹⁸/cm³) is on the top of the active region, followed by a 150-nm-thick p-type GaN contact layer (p doping = 1.0 × 10¹⁹/cm³). In order to study and analyze the optical properties of blue LED grown on patterned and planar sapphire substrates, the as-grown samples were characterized by high-resolution x-ray diffraction (HRXRD), atomic force microscopy (AFM), scanning electron microscopy (SEM) and Raman scattering. In order to reveal and obtain a comprehensive knowledge of the threading dislocation densities of Sample A and Sample B, we have measured X-ray rocking curves (XRCs) for both symmetry and asymmetry diffraction planes by HRXRD. The HRXRD was performed using a Bruker D8-discover system equipped with Ge (220) monochromator and channel-cut analyzer, delivering a pure Cu-Kα line of wavelength λ = 0.15406 nm; The AFM was performed using an Agilent 5500 scanning probe system and the micro-Raman measurements were carried out in backscattering geometry with the Raman spectrometer Jobin Yvon LabRam HR800. An argon laser of 514-nm wavelength was used as an excitation light source and a 50 × objective was employed to focus the incident laser light of a power of 14.2 mW on the sample, the spectrometer was calibrated using single-crystal silicon as a reference.

In order to study and compare the threading dislocation densities of Sample A and Sample B, the full-widths at half-maximum (FWHMs) of x-ray rocking curves were measured as shown in Fig. 2. The FWHMs of Sample A and Sample B are 483.23 arcsec and 305.98 arcsec for (002) planes, respectively, while for (102) planes, the respective FWHMs are 902.11 arcsec and 324.98 arcsec. It is well known that the FWHMs of XRD rocking curves from (002) planes or (102) planes are related to screw or edge dislocation densities [2, 8], and the narrower FWHMs of the XRD rocking curves are brought into correspondence with the lower threading dislocation density existing in the epifilms. Eq. (1) is used to be described as the relationships about threading dislocation density as following [9, 10].

in which ρ_s and ρ_e are the screw and edge threading dislocation densities, respectively; the quantities Δω_s and Δω_e refer to the FWHM of (002) and (102), respectively; a and c are the relevant Burgers vectors of GaN epifilms. The calculated results of threading dislocation densities existing in GaN epifilms on Sample A and Sample B are shown as Table 1. We can see that whether for screw dislocation and edge dislocation, the number dislocation density of Sample B is smaller than that of Sample A, which indicates that the screw or edge dislocation density existing in Sample B is lower than in Sample A. From an analysis of the threading dislocation density we are able to conclude that Sample B grown on patterned sapphire substrate has a better crystalline quality than Sample A grown on unpatterned planar sapphire substrate.

Figure 3 exhibits the SEM images and 5 × 5 μm AFM images of Sample A and Sample B. From Figs. 3(a) and 3(b), there appear regular hexagonal pyramids on the surface of Sample A and Sample B, and the number of hexagonal pyramids on the surface of Sample A is much larger than that on the surface of Sample B. It was well known that the hexagonal pyramid on the surface of epifilm is established when the islands are combined during the process of growth owing to the differences in lattice and thermal mismatch between sapphire and epitaxial epifilms. From HRXRD, lower dislocation density existing in Sample B corresponds with the smaller number of hexagonal pyramids on the surface of Sample B, indicating that the active region of LED grown on PSS has a better surface morphology, which coincides with the results of HRXRD analysis. Figs. 3(c) and 3(d) show the difference of surface fluctuation between Sample A and Sample B. A high percentage of the root mean square (RMS) on the measured area is between 65 nm and 75 nm for Sample A, and the RMS is between 35 nm and 45 nm for Sample B. In additional, the maximum RMS is 180 nm on the surface of Sample A, while the maximum RMS of Sample B is 65 nm, which shows that Sample B has a smaller dislocation density on surface morphology than Sample A does.

Raman scattering was performed to investigate residual strain of the two samples recorded in the z(xx)-z back-scattering mode with 514-nm wavelength excited from Ar⁺ at room temperature. The Raman shift is often used as a tool for strain assessment in mismatched epifilms. Raman scattering spectra of Sample A and Sample B are shown in Fig. 4 for a comparison. Raman peaks located at 572.14 cm⁻¹ and 569.14 cm⁻¹ are E₂ (high) phonon mode of GaN for Sample A and Sample B, respectively. It was well known that the blue-shift of GaN E₂ (high) phonon mode corresponds to compressive strain existing in epifims, which is caused by differences in the thermal expansion coefficient and lattice constant mismatches between the GaN epifilm and sapphire substrate [2, 11, 12]. As is well known, the lattice constant and thermal expansion coefficient mismatch between substrate and epitaxial film, characteristic for the heteroepitaxial growth of nitrides on foreign substrates, results in the presence of strain condition under linear elasticity biaxial system [1, 13]. The theory value of E₂ (high) for strain-free GaN is 567.5 cm⁻¹ [3, 14, 15], so there existed blue-shift of E₂ (high) mode of 4.64 cm⁻¹ and 1.64 cm⁻¹ relative to the theory value of 567.5 cm⁻¹ for Sample A and Sample B, respectively, suggesting that both GaN epifilms are under residual compressive strain. However, the smaller Raman shift of E₂ (high) of patterned sapphire substrate indicates that the compressive strain existing in Sample A is larger than that in Sample B. Therefore, the LEDs grown on patterned sapphire substrate possess less residual stress than those grown on planar sapphire substrate.

The room temperature PL spectra of Sample A and Sample B are illustrated in Fig. 5 for a comparison. The oscillatory structure of PL spectrum for Sample B has an apparent energy periodicity, which is similar with that reported in Ref. [16]. There are two differences from Ref. [16]. One difference is 24 meV energy periodicity at room temperature instead of 92 meV at the temperature of 50 K reported in Ref. [16]; the other difference is that it is grown on PSS instead of SiC. As can be shown, the main peak is at 451.89 nm (2.74 eV) for Sample B and at 477.03 nm (2.60 eV) for Sample A, which is the corresponding band emission peak of In_XGa_1-XN epifilms and as already mentioned above. It is larger residual strain existing in Sample A than Sample B that caused the difference in shift of the main peak. The intensity of main peak for Sample B is stronger than that of Sample A. There are many tiny peaks distributed at both sides of the main peak, demonstrating that there are several emission centers existing in the two samples. Figure 6 displays CIE1931 chromaticity coordinate calculation of Sample A and Sample B according to room temperature PL. We know that lower residual stress existed in the active region induced lower built-in electric field, eliminating the red-shift of emission wavelength of LED. In this sense, it is therefore advantageous to weaken the influence of quantum-confined Stark effect (QCSE) and high-current injection droop-efficiency effect, and greatly improve the quantum efficiency when blue-light LED is grown on PSS. Based on the above analysis, we have a reason to believe that an active region with superior optical properties is obtained when LEDs are grown on the patterned sapphire substrate. Therefore, a conclusion is drawn that LED with superior luminescence properties with higher crystalline quality, lower dislocation density, less defects on the surface and residual strain are obtained when the LED is grown on a patterned sapphire substrate.

In this paper, we have focused on the optical properties of the active region of blue light LED grown on PSS. HRXRD, AFM, SEM, photoluminescence and Raman scattering have been employed to investigate crystal quality, surface morphology, residual strain and optical properties. Results indicate that crystal quality of the LED can be improved greatly, the threading dislocation density and residual stress existing in the active region have been reduced when LED is grown on PSS, which provides a method to improve the quantum efficiency and eliminate the high-current injection droop-efficiency effect.