1. Introduction

Gallium nitride (GaN)-based light-emitting diodes (LEDs) have demonstrated remarkable success in applications for solid-state lighting and back lighting of display. Material quality [1,2], light extraction efficiency [3,4], and metal-semiconductor ohmic contacts [5] have progressed considerably to improve the luminous efficiency of GaN-based LEDs. Novel methods of growing epitaxial layers on a nano-patterned sapphire substrate have also been applied to reduce dislocation density in these devices [6–8]. The configuration of indium gallium nitride (InGaN)/ GaN multiple quantum wells (MQWs) also plays a key role in the internal quantum efficiency (IQE) of LEDs. Spreading the carriers over a larger volume by employing either a double heterostructure or an MQW active region can achieve an increased IQE at higher current densities [9]. In an InGaN/GaN MQW configuration, carriers must effectively increase the concentration and uniformly spread across the quantum wells to improve IQE and suppress the efficiency droop of LEDs. It was proposed to facilitate the uniform distribution of the carrier from results of simulation and experiment by reducing the thickness of GaN barriers in InGaN/GaN MQWs [10–14]. Although experimental reports have indicated an efficiency improvement on LEDs with a thin GaN barrier under high current injection [13,14], they presented less efficiency in reality than LEDs with a thick GaN barrier under low current injection. The possible cause of the inconsistency between results of simulation and experiment under low current injection is considered to be the material degradation of InGaN/GaN MQWs with thin GaN barrier. Degradation of the thin-barrier InGaN/GaN MQWs might be caused by indium (In) surface segregation [15–19], cumulative compressive strain, and degradation of InGaN/GaN interface quality [20,21]. Strain-compensated structures, such as InGaN/aluminum gallium nitride (AlGaN) and InGaN/AlGaN/GaN MQWs, have been proposed to compensate the compressive strain at the InGaN well, which can effectively reduce the total strain energy of MQWs [22–24]. Saito et al. and Hwang et al. [25,26] used a similar concept by introducing an AlGaN inserting layer as a strain-compensated layer in InGaN/(In)GaN MQWs to enhance the light-emitting efficiency of green–yellow and red InGaN LEDs. In our study, we proposed an aluminum nitride (AlN) insert layer between InGaN and GaN to improve the opto-electrical characteristics of thin-barrier InGaN/GaN MQW LEDs. This work discusses the effects of a thin-barrier InGaN/AlN/GaN MQW LEDs on the electrical and optical properties of GaN-based LEDs and the fabrication process.

2. Experiments

Samples were grown on a 2-in (0001) patterned sapphire substrate (PSS) through a Veeco K465i 14 × 4” metalorganic chemical vapor deposition (MOCVD) system. Diameter, spacing, and height of the PSS periodic convex pattern were 2.8, 0.2, and 1.8 µm, respectively. Trimethylindium, trimethylgallium, trimethylaluminum, and ammonia were used in MOCVD epitaxy as source materials of In, Ga, Al, and N, respectively. Bicyclopentadienyl magnesium and silane were used as p- and n-type doping sources, respectively. The temperature of the reactor was first increased to 900 °C to grow a 10 nm-thick AlN nucleation layer for 50 sec on the PSS. The temperature was then raised to 1050 °C to grow a 2 µm-thick undoped GaN epitaxial layer for 45 min followed by the deposition of a 2 µm-thick n-GaN layer for 20 min. It was reduced to 830 °C for the deposition of a 20-pair InGaN (2 nm)/GaN (2 nm) superlattice. The growth time of InGaN and GaN layer of superlattice are 20 and 30 sec, respectively. Subsequently, a conventional 9-pair thick-barrier InGaN (3 nm)/GaN (10 nm) light-emitting MQW structure was produced via high–low temperature scheme. The growth temperatures of the InGaN well and GaN barrier layers were maintained at 730 °C and 880 °C, respectively. The growth time of the InGaN well and GaN barrier layers were 2.5 and 2.25 min, respectively. Aside from the conventional MQWs, 9-pair thin-barrier InGaN (3 nm)/GaN (6 nm) and thin-barrier InGaN (3 nm)/AlN/GaN (6 nm) MQWs with 10- and 20-sec-deposited AlN insert layers were also prepared. The growth time of 6 nm-thick GaN thin barrier is 1.75 min. The AlN insert layers of InGaN/AlN/GaN MQWs are located in the interface on top of the InGaN wells and bottom of GaN barriers. After four different MQW structures being prepared, the growth temperature was increased to 930 °C for the deposition of 40 nm-thick low-temperature Mg-doped GaN layer (LT p-GaN) for 4 min. After the growth of the LT p-GaN, a 30 nm-thick Mg-doped Al_0.15Ga_0.85N layer for 5 min and a 70 nm-thick p-GaN layer for 10 min were sequentially deposited on top of each sample to form the full structure of LEDs. LEDs with MQW structures of InGaN (3 nm)/GaN 10 nm), InGaN (3 nm)/GaN (6 nm), and InGaN (3 nm)/AlN/GaN (6 nm) MQWs with 10- and 20-sec-deposited AlN insert layers were called LED I, II, III, and IV, respectively. Subsequently, standard processing steps were performed to fabricate 229 × 711 µm LED chips with indium tin oxide for the upper contact. The current–voltage (I–V) characteristics of the fabricated LEDs were then measured using an HP-4156C semiconductor parameter analyzer with a limitation of 100 mA. Output powers and emission wavelength of the LEDs were measured using a calibrated integrating sphere and a spectrometer (Ocean Optics USB2000) at room temperature, respectively.

3. Results and discussions

Figure 1 presents the I–V characteristics of the LED samples. All LED samples show a forward voltage (V_f) of 100 mA in the range of 3.4–3.55 V. The thin-barrier MQWs without the AlN insert layers had a slight effect on the V_f of the LEDs. However, the LEDs with thin-barrier MQWs without the AlN insert layers has a larger forward leakage current than standard LEDs (LED I) in a bias range of 1.5–2.5 V, which might be due to a carrier recombination in the depletion region through defects [27]. Thin-barrier MQW LEDs (LED II) has the highest forward leakage current in the same bias range. The thin-barrier InGaN/AlN/GaN MQW LEDs with 10-sec-growth and 20-sec-growth AlN insert layers do help to reduce forward leakage current. Moreover, the effect becomes more significant for the 20-sec ones. The larger forward leakage current of LEDs II, III, and IV than that of LED I implies that the quality of the thin-barrier InGaN/GaN MQWs is worse than that of LED I. The AlN insert layers could improve the quality of thin-barrier InGaN/GaN MQWs and reduce the forward leakage current. In addition, LEDs with thin-barrier InGaN/GaN MQWs shows the largest reverse current of 8.6 × 10⁻⁶ mA at a reverse −15 V bias, which is near 2-order magnitude and larger than that of LED I. We consider that the increase in growth time of the AlN insert layers can effectively suppress the reverse current at −15 V, because the reverse current of LED IV is approximately one order less than that of LED II at −15 V as shown in Fig. 1. Thus, results of the reverse current indicate better quality of the thin-barrier InGaN/AlN/GaN MQWs.

 

Fig. 1. I-V characteristics of all LED samples, which were measured by HP4156C semiconductor parameter analyzer.

Download Full Size | PDF

Meanwhile, LEDs with thin-barrier InGaN/AlN/GaN MQWs have positive effects on the light output power. Figure 2 shows the current dependent light output power (L–I curves) and external quantum efficiency (EQE) of the LED samples. The 120-mA light output power of LEDs I, II, III, and IV are 35.5, 31.9, 32.3, and 35.3 mW with emission wavelength of 454.5, 457.5, 456.8, and 455.5 nm, respectively. LED I has the largest 120-mA light output power and maintains the largest light output power for injection current larger than 100 mA. LEDs with thin-barrier InGaN/GaN MQWs (LEDs II) have the least light output power for the injection currents due to the degradation of the thin-barrier InGaN/GaN MQWs. The 120-mA light output power of LED II can be improved by inserting AlN thin layers into thin-barrier InGaN/GaN MQWs and can be recovered to almost the same as the 120-mA light output power of LED I by introducing the 20-sec-growth AlN insert layers. Moreover, thin-barrier InGaN/AlN/GaN MQW LEDs with 20-sec-growth AlN insert layers (LED IV) shows larger light output power than that of LED I for an injection current of less than 100 mA. The improved light output power of LED IV is considered to be attributed that the 20-sec AlN insert layers improved the quality of thin-barrier InGaN/GaN MQWs. Figure 2 shows the current dependent light output power of LEDs II and III, and that the largest light output power at approximately 200 mA is lower than that of LEDs I and IV. Thus, LEDs II and III exhibited a severe efficiency droop in high injection current region.

 

Fig. 2. Characteristics of current dependent light output power and external quantum efficiency of all LED samples. The injection current was supplied by Keithley 2400 source meter.

Download Full Size | PDF

LED II shows the least current dependent EQE in Fig. 2. The current dependent EQE of LED II is slightly improved for an injection current larger than 100 mA by introducing a 10-sec AlN insert layers in thin-barrier InGaN/GaN MQWs of LED II. The current dependent EQE of LED III shows a considerable improvement in injection current of less than 50 mA compared with that of LED II. Based on IQE, η is proportional to B n² / (A n + B n² + C n³), where parameters A, B, and C represent the Shockley–Read–Hall (SRH) nonradiative, bimolecular radiative, and Auger recombination, respectively [28,29]; the SRH nonradiative recombination degrades the IQE of LEDs under low current injection. Consequently, the EQE improvement of LED III under low injection current indicates less SRH recombination than that of LED II because of the 10-sec-growth AlN insert layers in thin-barrier InGaN/GaN MQWs. LED IV presented a significant EQE improvement for the injection current by increasing growth time of AlN insert layers from 10 s to 20 s, particularly for low injection current, compared with LED II. The EQE of LED IV is still smaller than that of LED I for an injection current larger than 100 mA. However, LED IV presented the largest EQE peak, and the EQE of LED IV is larger than that of LED I for an injection current of less than 100 mA. Therefore, the SRH recombination rate of LED IV is less than that of LED I. The thin-barrier InGaN/AlN/GaN MQWs with 20-sec-growth AlN insert layers might have better quality than standard thick-barrier InGaN/GaN MQWs. However, all thin-barrier InGaN/GaN and InGaN/AlN/GaN MQW LEDs indicated a larger efficiency droop than LED I. Efficiency droop is defined by EQE differences between the peaks of EQE. EQE at 250 mA over peak EQE of LEDs I, II, III, and IV were 36.3%, 49.5%, 51.7%, and 45.1%, respectively.

The mechanisms of carrier recombination can be deduced by approximating the L–I curves of LEDs as the power law, that is, L is proportional to I^S, where the power of S presents the dominant carrier recombination process in the light emission process [30,31]. Kim et al. [32] reported that the power of S could be extracted from the derivative of Log L with respect to Log I. Figure 3 presents the current dependent power of S (S–I curves), and the insert in Fig. 3 is the enlarged S–I curves for current less than 20 mA. LED II showed the largest S value, which was approaching the value of 2, for current less than 10 mA. The increment of S value with reduction of current is steeper than the rest of the LEDs. The SRH recombination will become dominant effects for injection current less than 10 mA for LED II because the S value is close to 2 [32]. Figure 4 shows the X-ray diffraction (XRD) spectra of LED I, II, and IV. The period of satellite peaks of thick-barrier InGaN/GaN MQWs of LED I is less than that of thin-barrier InGaN/GaN and InGaN/AlN/GaN MQWs of LED II and III. According to the periods of satellite peaks of XRD spectra of LED I, II, and IV the corresponding well-barrier pair thicknesses of thick-barrier InGaN/GaN MQWs, thin-barrier InGaN/GaN MQWs, and thin-barrier InGaN/AlN/GaN MQWs with 20-sec-growth AlN are around 13.0, 8.9, and 9.2 nm, respectively. The thin-barrier InGaN/GaN MQW LEDs presents larger average In composition than the thick-barrier InGaN/GaN MQW LEDs because the thickness ratio of InGaN well/GaN barrier is larger for thin-barrier InGaN/GaN MQWs. The thin-barrier nine-period InGaN/GaN MQWs might have larger strain than thick-barrier nine-period InGaN/GaN MQWs because of higher average In composition of thin-barrier InGaN/GaN MQWs. The enlarged strain of thin-barrier nine-period InGaN/GaN MQWs might result in defects, such as In surface segregation, degradation of InGaN/GaN interface quality, and roughened InGaN/GaN morphology. The thin-barrier

 

Fig. 3. Current dependent power of S (S-I curves) and (inset) enlarged S-I curves of all LED samples for current less than 20 mA.

Download Full Size | PDF

 

Fig. 4. The X-ray diffraction (XRD) spectra of LED I, II, and IV.

Download Full Size | PDF

InGaN/AlN/GaN MQWs with 20-sec-growth AlN insert layers shows less average In composition than the thin-barrier InGaN/GaN. Although the estimating thickness of 20-sec-growth AlN insert layers in MQWs is around a monolayer, the AlN insert layers in thin-barrier InGaN/GaN MQWs would compensate the strain because the lattice constant of AlN is a little less than those of GaN and InGaN that will provide opposite strain. And LED IV seems presenting clearer XRD spectrum peaks and shape than LED II. Consequently, the thin-barrier InGaN/AlN/GaN MQW LEDs with 20-sec-growth AlN insert layers maintain their S value close to 1 to make an injection current as low as 4 mA. Thus, the radiative recombination is dominant until the injection current of 4 mA [32]. S values of the LEDs were less than 1 in the high current region. Nevertheless, LEDs II and III presented smaller S values and a faster drop in S value with increasing current than LEDs I and IV in high current range, indicating that LEDs II and III might have a more severe carrier overflow than LEDs I and IV in the high current range [32]. The strain-induced quality degradation of thin-barrier InGaN/GaN MQWs might affect magnesium (Mg) doped AlGaN current blocking layer (EBL) of LEDs. There is a clear shoulder peak on the right side of GaN main peak of XRD spectrum of LED I that should be from Mg doped AlGaN EBL. But the Mg doped AlGaN EBL shoulder peak of XRD spectra of LED II and IV is not as clear as that of LED I form Fig. 4. And it might indicate the change of Al composition of Mg doped AlGaN EBL. Studies have reported that the strain type of layers underneath an AlGaN layer might influence the Al composition and its inhomogeneity in the AlGaN layer [33,34]. The high strain of thin-barrier InGaN/GaN MQWs might reduce the Al composition and enhance the inhomogeneous Al composition of AlGaN EBL, resulting in the severe carrier overflow of LEDs in the high current range. The strain-compensated thin-barrier InGaN/AlN/InGaN MQWs will relieve the afore mentioned effects of AlGaN EBL to suppress the carrier overflow of LEDs in the high current range and improve the efficiency droop of LEDs. However, the preceding discussion is only speculative, and further research is needed to validate it.

4. Conclusion

In conclusion, the thin-barrier InGaN/GaN LEDs did not show better opto-electrical properties, such as light output power, than traditional thick-barrier InGaN/GaN MQW LEDs. The larger strain of thin-barrier InGaN/GaN MQWs than that of traditional InGaN/GaN MQWs might have a negative effect on the quality of thin-barrier InGaN/GaN MQWs. Fortunately, the strain of thin-barrier InGaN/GaN MQWs could be compensated by introducing AlN insert layers to form strain-compensated thin-barrier InGaN/AlN/GaN MQWs. The AlN insert layers in MQWs would significantly improve the opto-electrical properties of thin-barrier InGaN/GaN LEDs, such as I–V characteristics, light output power, and efficiency droop, thereby remarkably suppressing the nonradiative recombination of LEDs in the low current range. The 120-mA light output power of thin-barrier InGaN/GaN MQW LEDs could be enhanced from 31.9 mW to 35.3 mW by introducing the 20-sec-growth deposited AlN insert layers, reaching almost the same light output power (120 mA) as traditional thick-barrier InGaN/GaN MQWs. Furthermore, the thin-barrier InGaN/AlN/GaN MQW LEDs with 20-sec-growth AlN insert layers also shows the largest peak EQE, and high efficiency in low current injection. Although thin-barrier InGaN/GaN MQW LEDs has a severe carrier overflow effect, which degrading the LED performance in high current injection. Carrier overflow of thin-barrier LEDs in high current injection can be suppressed by introducing thin-barrier InGaN/AlN/GaN MQW with 20-sec-growth AlN insert layers.