1. Introduction
Semiconductor laser diodes (LDs) have tremendous influence on human's life, since they have been applied in many field such as optical fiber communication, optical data storage in the past decades. Laser display using red, green and blue (RGB) LDs as light source is an emerging display technology as it shows larger color gamut, higher color saturation, and capability for both pico-projector and larger size display. Figure 1 shows the color gamut of laser display versus other display technology. Pico-projector and televisions have been considered to be the most important market for the laser display system[1-5]. The performance of green LDs used in pico-projectors should meet the following specifications: wavelength in the range of 520 to 532 nm, a wall plug efficiency (WPE) of at least 5% at 65 ℃, reliability over more than 10 000 h and an output power exceeding 80 mW[5]. While more than 10 W light power is needed for each color for laser television with 100 inch size.
Direct blue and green emission semiconductor LDs made of GaN-based materials are desirable for laser display. However, fabrication of GaN-based green LDs is more challenging, compared to that of blue and ultraviolet LDs[6, 7]. Peak gain of GaN-based LDs decreases as emission wavelength increases from violet to green range, which results into increasing threshold current and decreasing efficiency of green LDs[8]. As early as 2001, researchers from Nichia Corporation had demonstrated InGaN multi-quantum-well structure laser diodes (MQW-LDs) with an emission wavelength of longer than 420 nm[9], but it was not until 2009 that researchers of Nichia Corp. demonstrated green (λ >500 nm) InGaN-based lasers grown on c-plane self-standing GaN substrates with lifetimes of 5000 h[10]. Nevertheless, due to the promising application in laser display, development of green LDs has attracted considerable attention of researchers, e.g. Nichia, Osram, Suzhou Institute of Nano-Tech and Nano-Bionics (Sinano), University of Michigan (UM), Sumitomo & Sony (SS), University of California Santa Barbara (UCSB) and so on. Different crystallographic orientation has been used by them as Figure 2 shows and the details of their work will be discussed later. In this paper, challenges, recent progress and outlook of green LDs will be discussed.
2. Challenges
Optoelectronic devices based on either GaN or GaAs suffer from a steep drop of efficiency when the emission wavelength is approaching the green region[11], which is known as the "green gap". We are facing lots of challenges to develop green LDs with good performances.
2.1 Crystalline defects
No matter which orientation (c-plane,a-plane,m-plane or r-plane) is used, the first challenge is the preparation of high-quality In-rich InGaN active layers in green LDs. Around 30% In needs to be incorporated in InGaN quantum wells (QWs) to realize green LDs, which results into the formation of crystalline defects in three aspects. Firstly, as a result of weak chemical bond of In-N, the growth temperature for InGaN needs to be lowered to incorporated enough In. Consequently, both the atomic diffusion length and the decomposition efficiency of NH3 decrease at low temperature, which results into crystalline defects, fluctuation of In composition, and rough interface in InGaN QWs because growth temperature needs to be high enough to maintain a step-flow growth which is essential for a high-quality epitaxial growth[12]. Fluctuation of In composition and rough interface in InGaN QWs significantly broaden the gain spectra of green LDs[13], as shown in Figure 3. Secondly, the ideal growth temperature for p-AlGaN cladding layer is as high as 1000 ℃ to ensure low resistance. However, such high temperature is going to induce serious thermal degradation of InGaN QWs[1]. Thirdly, lattice mismatch between InGaN QWs and GaN is as large as 3.3%, which induces crystalline defects in InGaN QWs. Crystalline defects not only reduce the radiative efficiency[14], but also shorten device lifetimes[15].

2.2 Quantum confined Stark effect
GaN-based materials with c-plane orientation are the most mature in terms of epitaxial growth. However, quantum confined Stark effect (QCSE) is a challenge to fabricate c-plane green LDs. Many literatures[16-21] have detailed reports about QCSE both experimentally and theoretically. QCSE originates from the polarization electric fields caused by both spontaneous and piezoelectric polarization charges at heterointerface [22, 23]. The polarization electric field is estimated to be more than 2 MV/cm in green InGaN QWs[24, 25]. Consequently, the overlap of e-h wavefunction is greatly reduced, as shown in Figure 4(b), which results into low luminescence efficiency.
Since c-plane InGaN QWs suffer from the QCSE, nonpolar (a-plane,m-plane) InGaN QWs free from QCSE and semipolar (r-plane) InGaN QWs with small QCSE are considered to have advantages to fabricate high performance green LDs[26]. Some groups (SS[27,28], UCSB[29,30]) have demonstrated their achievements in green LDs based on these planes. However, crystalline defects such as stacking fault[31] and misfit dislocations readily appear in nonpolar and semipolar InGaN QWs, which limit the application of those planes.


2.3 Optical confinement
Since the refractive index depends on wavelength, with increasing wavelength, the refractive index difference between GaN waveguide layer and AlGaN cladding layer decreases. As a result, there will be a leakage into substrate as shown in Figure 5, based on the simulation and measurement for the vertical near and far field of a given laser structure (emitting at 497 nm) demonstrated by Lermer et al.[32]. We[33] have confirmed that confinement factors of blue and green LDs are smaller with respect to that of violet LD when their device structure is similar and the results are shown in Figure 6. Particularly, the influence of ridge height, p-AlxGa1-xN composition and thickness on confinement factor and absorption loss are analyzed based on the simulation and experiment. Increasing the thickness of AlGaN, substituting InGaN for GaN as the waveguide both can suppress the leakage of light, while the lattice mismatch will be aggravated at the same time. Therefore, QCSE induced by strain may become more serious and even crack will be found in epitaxy layer, hence we are supposed to optimize the design of green LDs' structure theoretically and experimentally.
2.4 Carrier leakage
With the increasing of injection current, the slope efficiency of green LDs was found to decrease. This phenomenon is attributed to the reduction of the injection efficiency induced by carrier leakage. The electron blocking layer (EBL) has been employed to reduce the electrons leakage successfully[34]. Nevertheless, there are limited studies on hole transport properties especially at high current density. Controversial results have been reported based on different model[35-37], the details are discussed in Section 3.4.
3. Progress
3.1 Two-temperature growth of QW active region
Since the growth temperature of green InGaN QWs has to be lowered to incorporate enough In, we employ a two-temperature growth technique to grow InGaN/GaN QWs, in which the growth temperature is raised to grow GaN quantum barriers (QBs). In a two-temperature growth process, InGaN QW tends to locally decompose during temperature ramping up to QB temperature in the view of low thermal stability of InGaN, which results into rough interface and In composition fluctuation. To prevent the InGaN from decomposing as growing the GaN, a GaN cap layer is grown on top of the QW, which is shown in Figure 7. The effects of GaN cap layer thickness have been studied by us[38]. The EL full width at half maximum (FWHM) is used to characterize the extent of inhomogenous broadening and potential homogeneity in the active region. As shown in Figure 8(a), the dashed line show the tendency that the EL FWHMs of LD increased with emission wavelength, which is an indication of enhanced potential inhomogeneity as indium composition in the InGaN QWs increases[1]. The sample with 1.8 nm LT-cap shows a similar slope compared with the 2.5 nm LT-cap, which means the thickness of 1.8 nm and higher is fully enough to protect the QW from temperature ramping.

As illustrated in Figure 8(b), the EL FWHM of our LD is compared with value from other groups. It should be noted that the data is dependent on the measured condition in different groups. The current density of measurement from Sinano' LDs is 10 A/cm2 under DC condition, while it is 14 A/cm2 for semipolar green LEDs reported by UCSB[39,40],150 A/cm2 for semipolar green LD structures reported by Sumitomo Electric Industries[41], PL Data are used for c-plane green LD structures of Osram[42] and Nichia[43]. It shows that FWHM value for emission wavelength of 530 nm (2.34 eV) is roughly comparable to that from others. In order to study the effect of LT-cap thickness on the microstructures of InGaN/GaN active region, HAADF-STEM measurement were carried out, as Figure 9 shows. QWs with 1.2 nm LT-cap are broken and separated InGaN islands can be observed clearly, which means the 1.2 nm LT-cap is not thick enough to protect the QW from being damaged. In contrast, for the LD structures with 2.5 nm LT-cap layer, QWs do not suffer from the temperature ramping.

3.2 Suppression of thermal degradation
Reference [44] has reported that the solubility of In in GaN is calculated less than 6% at 800 ℃. As result, with the increase of indium concentration, the thermal stability of InGaN QW decreases. Decomposition of InGaN and formation of In clusters will appear in the active region during p-type layer growth, which is called the thermal degradation. This problem is particularly severe in laser diode epitaxial growth since longer growth time and higher growth temperature are used to grow p-AlGaN/GaN SL cladding layer. As shown in Figure 10(a), there are high density of dark regions with different gray scales which indicates no or weak emission is detected in micro-PL images of LD caused by the thermal degradation. The different gray scales may indicate the decomposition in the top and bottom QW doesn't happen simultaneously. The less dark area marked with arrow "1" in Figure 10(a), may imply that in this region only one InGaN QW layer is degraded, but another QW layer still performs well. For the area marked with arrow "2", both top and bottom InGaN QW layers are damaged resulting in a much weaker emission.

To mitigate the thermal budget to InGaN QWs, the growth temperature of p-type layers was lowered by 30 ℃ to grow LD-Ⅱ sample. Figure 10(b) shows the micro-PL image of LD-Ⅱ. Compared to LD-I, it is noted that both the size and the density of the dark spots are reduced, indicating that the decomposition of InGaN QW is suppressed to certain degree, which results from reduced thermal budget to InGaN QWs during p-type growth. In order to fully suppress the decomposition of InGaN QWs, preventing the formation of In-rich clusters is critical. Several previous works reported that growth interruption[45] or introduction of H2 after the growth of InGaN well layer[46,47] can remove In-rich InGaN clusters on the InGaN QW upper interface and enhance the thermal stability of InGaN well layer. However these approaches may blueshift the emission wavelength greatly which needs to be avoided when growing green InGaN LDs. Here we reduce the thickness of the LT cap layer which is employed to protect InGaN QW layer from evaporation during the subsequent temperature ramping process. A suitably thin LT-cap layer allows a slight evaporation of the InGaN surface without shortening the emission wavelength remarkably. Figure 10(c) shows the micro-PL image of LD-Ⅲ which has thinner LT cap layer compared to LD-Ⅱ. It is noted that the luminescence is extremely homogeneous and the dark regions disappear totally. As a result of suppression of local InGaN decomposition and improvement of luminescence homogeneity, the EL intensity of LD-Ⅱ and LD-Ⅲ are greatly enhanced by 110% and 450% compared with LD-I, as shown in Figure 10(d).
3.3 Carbon impurity in p-AlGaN:Mg cladding layer
In order to suppress the thermal degradation of high indium content green InGaN/GaN QWs during the growth of p-type AlGaN cladding layers[1,48], p-type cladding layers are usually grown at a temperature lower than the optimal temperature which is higher than 1000 ℃ for AlGaN. However, p-type AlGaN grown at a low temperature often shows high resistivity due to increased defects & impurity incorporations[49-52]. Because of the chemical activity of Al, AlGaN layer grown by MOCVD usually contains higher carbon impurity concentration than GaN especially when the growth temperature is reduced than the optimal growth temperatures. We[53] have studied the dependence of carbon concentration on growth conditions. As shown in Figures 11(a),11(b) and 11(c), when the p-type AlGaN cladding layer is grown at a lower temperature to avoid the thermal degradation of the active region, the incorporation of carbon impurity will be enhanced. For example, simply reducing the growth temperature of AlGaN:Mg to 907 ℃ leads to an increase of carbon concentration to as high as 4 × 1018 cm-3. However, by adopting a lower growth rate and a higher growth pressure, p-type AlGaN with carbon concentrations less than 1× 1017 cm-3 can be achieved We have observed a correlation between carbon concentration and electrical properties of AlGaN:Mg. When the carbon concentration is higher than a certain level in the order of 1E17, there is a clear tendency that the hole concentration decrease and thus the resistivity increases as the carbon concentration increases in the samples, as shown in Fig. 11(d). The mechanism behind this was discussed in our previous paper[54].
3.4 Carrier leakage
Injection efficiency is one of the main factors which affect LD performance. Compared to that of electron, the effective mass[55] of hole is much heavier and the mobility[56,57] of hole is significantly lower. The severe asymmetry in carrier transport results into that holes are mainly confined in the quantum well (QW) nearest the p-side at low current density[58,59], while electron leakage[60-65] occurs easily even at low injection level of 20 A/cm2 typical for LED operation. Employing and optimizing the electron blocking layer[66-71] is an effective method to reduce electron leakage in Ⅲ-nitride LDs. On the other hand, there are limited studies on hole transport properties at high current density for LD operation (more than 1000 A/cm2, and controversy results were reported. According to the reports of Sizov et al.[72,73], piezoelectric field hinders hole transport and holes are mainly confined in the QW nearest the p-side in the c-plane InGaN-based green LDs. However, according to the reports of Hager et al.[74,75], holes can overflow from the green active region in the c-plane green LDs and recombine in the n-InGaN layer right below green active region. Zhang et al.[76] report that large amounts of holes can penetrate through seven In0.4Ga0.6N/GaN QWs with the assistance of piezoelectric field at high current density.
We have investigated the hole transport in c-plane InGaN-based green LDs by both simulations and experiments[36]. It is found that holes can overflow from the green double quantum wells (DQWs) at high current density by using a specially designed green LD structure, as shown in Figure 12(a). The active region contains four blue InGaN/GaN QWs that are used as detection layer, hence blue luminescence is supposed to be observed if the holes are transported to the region near n-side. According to the EL spectra shown in Figure 12(b), it is confirmed that holes can overflow from the green DQWs. And about 30.7% of holes recombine outside the DQWs at 1000 A/cm2 which is based on the simulation shown in Figure 12(c). It is also found that a heavily silicon-doped layer right below the green DQWs can suppress hole overflow from the green DQWs which is shown in Figure 12(d), the intensity of blue luminescence almost disappears.
3.5 Summary
Although there are so many challenges to fabricate green LDs, considerable progress has been made as we mentioned. Green LDs (λ >500 nm) have been achieved by several groups and companies such as Osram, UM, UCSB, SS, Soraa, Corning, Rohm and Sinano listed in Table 1. Nichia and Osram are leading the progresses in the development of green LDs. Both of them fabricated green LDs based on c-plane. Nichia reported high-power green LD (Pout>1 W,λ =525 nm, Jth=1.68 kA/cm2 , which shows potential to be used in large-size laser TV. Sinano has demonstrated green LDs grown on bulk GaN substrate which lases at 508 nm as shown in Figure 13(a). Figure 13(b) shows the power-current-voltage curve of a typical green LD under continuous-wave operation at room temperature. It can be shown that the threshold current density and the threshold voltage is 1.8 kA/cm2 and 4.4 V, respectively. The output power is 58 mW at 6 kA/cm2.
![]() |