Performance characteristics of violet, blue, and green (0001) c-plane InGaN quantum well LEDs and laser diodes
(0001) c-plane InGaN light-emitting diodes (LEDs) already exhibit outstanding levels of performance. This, of course, can also be attributed to two decades of intense research worldwide, since the first demonstration of blue III-nitride- based LEDs on (0001) GaN/sapphire by Nakamura et al. (1991, 1994, 1995). Figure 8.2 shows the measured external quantum efficiencies (EQE) of InGaN single and multiple quantum well LEDs grown on c-plane (0001) sapphire substrates emitting in the violet-to-green spectral range. In addition, the EQEs of InGaP-based LEDs emitting in the red, orange, and yellow wavelength spectrum are plotted. As can be seen from Fig. 8.2 the EQE of InGaN-based LED peaks near 450 nm and rapidly decreases for longer wavelengths. Although yellow and even red-emitting c-plane InGaN QW LEDs have been demonstrated (Mukai et al., 1999), their efficiencies cannot compete with the performance of InGaP-based red and yellow LEDs. However, the peak efficiencies of InGaP LEDs appear also to be limited to a narrow spectral range near 650 nm, and is rapidly decreasing for shorter wavelengths. This situation creates a spectral
Fig. 8.2. Reported external quantum efficiencies (EQEs) for c-plane, non-polar, and semipolar InGaN quantum-well LEDs as well InGaP-based LEDs at different emission wavelengths (Morita et al., 2004; Nichia Corporation, 2012a).
region in the green-yellow wavelength range between about 530 nm and 600 nm, where all semiconductor LEDs exhibit relatively poor external quantum efficiencies—a circumstance that is often addressed as the so-called “green gap”. For InGaN quantum-well LEDs the drop in EQE originates from multiple causes: (1) at longer wavelengths the increasing indium mole fraction in the InGaN QWs results in an increase in the spontaneous and piezoelectric polarization charges at the InGaN/GaN interfaces (Schmidt et al., 2007). The resultant polarization fields result in a spatial separation of the electron and hole envelope wave-functions and a red-shift of the emission wavelength due to the quantum-confined Stark effect (QCSE) (Miller et al., 1984). Simulations of the electron-hole wavefunction overlap for (0001) InGaN/GaN quantum well show that the envelope electron-hole wavefunction overlap is reduced by more than 50 % when going from a blue emitter to a green or yellow LED (Arif et al., 2008). Consequently, the radiative recombination lifetimes in the InGaN QWs increase significantly at longer emission wavelengths (Chichibu et al., 1999). (2) In addition, at higher indium mole fraction the formation of extended and point defects in the InGaN layers is enhanced, leading to shorter non-radiative recombination lifetimes (Lu et al., 2010; Chichibu et al., 2006). Since the internal quantum efficiency niQE depends on the radiative lifetime Trad as well as non-radiative lifetime Tnr according to following relationship
both of these detrimental effects will lead to a reduction in the internal and consequently also external quantum efficiency of the light-emitting devices. Obviously, the effects also depend on additional factors, such as the threading dislocation density and temperatures (Karpov and Makarov, 2002; Chichibu et al., 2006) and may be partially suppressed by potential barriers forming threading dislocations (Hangleiter et al., 2005).
Furthermore, the EQE also changes with the injection current and current densities—a phenomenon that is also often described as “droop” or “efficiency droop”. As can be seen in Fig. 8.3, droop is observed for GaN-based LEDs at all emission wavelengths starting at different current densities. Droop is a significant impediment for high-power LEDs, e.g. for application in solid-state lighting, where high light output power levels at low chip cost are required, which can be achieved only by operating these devices at high current densities. For example, the first generation of high-power InGaN MQW LEDs were operated at a dc current of 350 mA with chip sizes in the order of 1 mm2. Currently, high- power LEDs are operating at currents up to 2 A, which corresponds to current densities of 200A/cm2 for the same chip size, and future generations of LEDs will certainly be driven at even higher current levels. The physical mechanisms underlying the droop effect are still not completely understood, and are intensively discussed among a number of research groups. Proposed mechanisms for the droop effect include Auger recombination (Shen et al., 2007; David and
Fig. 8.3. (a) Emission spectra for blue, cyan, and green (0001) c-plane InGaN MQW LEDs measured at a forward current of 20 mA. (b) Measured external quantum efficiencies for blue, cyan, and green (0001) c-plane InGaN MQW LEDs vs. dc forward current. Courtesy of C. Reich (Technical University, Berlin).
Grundmann, 2010; Laubsch et al., 2009; Kioupakis et al., 2011), carrier leakage (Kim et al., 2007b; Piprek, 2010), density-activated defect recombination (DADR) (Hader et al., 2010), and changing radiative recombination rate and rate equation analysis (Ryu et al., 2009). It is not the intention here to discuss the pros and cons of the different possible mechanisms. However, almost regardless of the specific cause, the droop effect is certainly influenced by the strong polarization fields in c-plane light-emitting heterostructures. For example, the large polarization fields in InGaN/GaN QWs limit the thickness of the QWs, since thicker QWs will result in a stronger separation of electron and hole envelope wavefunctions, due to the QCSE. Thick InGaN QWs, however, are preferred for reducing the carrier density in the QWs, in order to reduce Auger recombination and carrier leakage-related droop effects. Therefore, reducing or eliminating the polarization fields in InGaN QWs should also lead to a reduced droop in III-nitride-based LEDs.
Another challenge for (0001) c-plane light-emitters is the change in peak emission wavelength with increasing current density. The high carrier densities at high current densities lead to a partial screening of the polarization field in the InGaN QWs, which in turn reduces the quantum-confined Stark effect. As a consequence, the peak emission is shifted to shorter wavelengths at higher current densities (Mukai et al., 1999). This effect is more pronounced for longer- wavelength LEDs, where the higher indium mole fraction in the InGaN QWs leads to stronger polarization fields. Figure 8.3 shows the emission spectra and external quantum efficiencies for blue and green c-plane InGaN MQW LEDs measured at different current densities, and Fig. 8.4 shows the change in emission
Fig. 8.4. Change in peak emission wavelength vs. dc forward current for blue, cyan, and green (0001) c-plane InGaN MQW LEDs. Courtesy of C. Reich (Technical University, Berlin).
wavelength for the same set of LED vs. the drive current. As can be seen, in all cases the emission shifts to shorter wavelengths at higher currents due to the screening of the polarization fields with higher carrier densities. It should be noted that part of the wavelength shift can be attributed to the filling of potential fluctuations that originate from compositional inhomogeneities in the InGaN QWs. However, this alone cannot explain the observed behavior.
Similarly, c-plane InGaN QW laser diodes are affected by the polarization fields. Although the first violet laser diodes were demonstrated in late 1995 (Nakamura et al., 1996a; Nakamura et al., 19966; Nakamura and Fasol, 1997) and continuous-wave operation with low threshold current densities were shown just a few years later (Nakamura et al., 1998a; Nakamura et al., 19986), the development of blue- and especially green-wavelength InGaN laser diodes lagged significantly compared to the rapid development of high-efficiency blue and green LEDs. For many years the realization of green InGaN-based laser diodes was considered extremely difficult, if not impossible. Although many factors contribute to these challenges, the large polarization fields in c-plane InGaN QWs were considered to be one of the main causes for the great challenges to realize green laser diodes. Therefore, the development of laser diodes on non-polar and semipolar GaN surfaces is considered one of the most promising approaches for the realization of high-efficiency and high-power true green laser diodes.
Nevertheless, by advancing the waveguide heterostructure and reducing nano- scopic and microscopic defects in the InGaN materials, several groups have
Fig. 8.5. Reported threshold current densities of (0001) c-plane InGaN QW laser diodes for different emission wavelengths. The different sets of data also represent the different time-periods in which the data have been obtained (Kozaki et al., 2007; Kim et al., 2008; Queren et al., 2009; Avramescu et al., 2009; Miyoshi et al., 2009; Nichia Corporation et al., 20126; Avramescu et al., 2010; Lutgen et al., 2010; Lermer et al., 2010; Schwarz and Scheibenzuber, 2011; Muller et al., 2011).
been able to push the emission wavelength of c-plane InGaN QW lasers well into the green spectral range. Figure 8.5 shows the reported threshold current densities of c-plane InGaN QW laser diodes for different emission wavelengths reported for different time-periods. Currently c-plane green laser diodes emitting at wavelengths as long as 532 nm have been demonstrated (Schwarz and Scheibenzuber, 2011). At slightly shorter wavelengths, cw operation of ridge- waveguide laser diodes with output power levels of more than 100 mW have been shown (Schwarz and Scheibenzuber, 2011). Almost simultaneously, high-power green laser diodes on different semipolar GaN surfaces have been demonstrated by several groups (Kubota et al., 2008; Okamoto et al., 2009; Yoshizumi et al., 2009; Enya et al., 2009; Kelchner et al., 2009; Tyagi et al., 2010; Raring et al., 2010; Yanashima et al., 2012) emitting at wavelengths as long as 535nm, with a light output of more than 100 mW. At the moment the performance characteristics of green lasers on polar and semipolar GaN are very close and which of these approaches will succeed will be determined sometime in the future. Parameters such as threshold current densities, wallplug efficiency, and output power will play an important role in that decision, but other important criteria, such as laser diode lifetimes, manufacturability, and chip cost will certainly play an important role as well.