1. Introduction

InGaN-based long wavelength laser diodes (LDs) have attracted considerable attention in the past few decades due to the wide potential applications in laser display, solid-state lighting, and undersea wireless communication [1–5]. Nowadays, most of the InGaN-based LDs are directly homo-epitaxially grown on small-size and costly free-standing GaN substrates, which makes the LD chip cost 2-3 orders of magnitude higher than that of light-emitting diodes grown on hetero-epitaxial substrates [6]. The utilization of large-diameter and cost-effective silicon (Si) substrates to grow InGaN-based LDs has been preferred to realize the on-chip light sources for Si photonics, which can further boost the application of LDs [7–11]. By virtue of the carefully engineered Al composition step-graded AlGaN multiple-layers, the compressive strain can be built up to compensate the thermal tensile stress and facilitate the bending and annihilation of threading dislocations (TDs), and therefore overcomes the major obstacles in growing crack-free and high-quality GaN on Si due to the huge mismatch in both lattice constant (∼17%) and coefficient of thermal expansion (∼54%) between GaN and Si [12–14]. Besides, the InGaN/GaN composite waveguides and AlGaN cladding layers are typically used for both optical confinement and strain control [15], which constitutes the Fabry-Perot resonant cavity for InGaN-based LDs.

We have previously successfully demonstrated the InGaN-based violet (415 nm) and blue (450 nm) LDs grown on Si substrate [8, 9]. To further exploit the potential of InGaN-based LDs for solid-state lighting and laser displays, InGaN-based LDs grown on Si emitting towards green wavelength are highly desirable. However, long-wavelength emission requires high In-content InGaN multi-quantum wells (MQWs) which faces a large lattice mismatch between GaN and InGaN, resulting in severe degradation of InGaN MQWs including the phase separation and interface roughness [16–18]. Moreover, the difference in refractive index between III-nitride materials becomes smaller as the lasing wavelength increases which actually reduces the optical confinement and increases the optical loss [19]. Besides, InGaN lower waveguide (LWG) layer grown with a relatively low temperature may induce the V-pit defects spreading across the InGaN LWG and MQWs, which not only leads to substantial non-uniform injection of carriers but also increases the photon scattering at interfaces and finally exacerbates the internal absorption loss of LDs [20–22]. These challenges prevent the lasing wavelength of InGaN-based LDs grown on Si from extending. Therefore, proper interface engineering to improve the quality of high In-content InGaN MQWs while promoting the optical confinement at the same time for InGaN-based long wavelength LDs grown on Si is urgently required.

In this work, we demonstrated a current-injected 480 nm InGaN-based cyan LD directly grown on Si substrate under a pulse operation at room temperature. A smooth morphology with trace amounts of V-pits of InGaN LWG layer and sharp interfaces of the InGaN MQWs were obtained by using a graded-compositional InGaN LWG capped with a 10-nm-thick Al_0.1Ga_0.9N layer. Moreover, the optical simulation results indicate that the optical confinement factor of InGaN active region can be enhanced as well due to the introduction of graded-compositional InGaN LWG layer.

2. Experiments

The epitaxial growth of InGaN-based LD was performed on Si(111) substrates by metal organic chemical vapor deposition. The schematic diagram of LD can be found in Fig. 1. The epitaxial structure consists of a 300-nm-thick AlN nucleation layer, Al composition step-graded AlGaN multiple-layers including an 170-nm-thick Al_0.35Ga_0.65N layer and a 310-nm-thick Al_0.17Ga_0.83N layer, a 3-μm-thick n-type GaN layer, an 1.2-μm-thick n-type Al_0.075Ga_0.925N lower optical cladding layer, an 160-nm-thick graded-compositional InGaN LWG layer with a 10-nm-thick undoped Al_0.1Ga_0.9N interlayer, two pairs of InGaN MQWs consisting of In_0.25Ga_0.75N QW, GaN/Al_0.07Ga_0.93N/GaN multiple-layer quantum barriers (QBs), an 140-nm-thick In_0.02Ga_0.98N upper waveguide layer, a 20-nm-thick p-type Al_0.2Ga_0.8N electron blocking layer, 600-nm-thick p-type Al_0.13Ga_0.87N/GaN superlattices upper optical cladding layers, and finally a 30-nm-thick heavily doped p-type GaN contact layer. The crystalline quality of GaN film grown on Si was evaluated by double crystal X-ray rocking curves and the density of threading dislocations was estimated to be 6.0 × 10⁸cm^-2. The graded-compositional InGaN LWG layer was grown with the In-content monotonically increasing from 0 to 6% to ensure an average In-content of 3%, followed by an 1-nm-thick GaN cap layer.

 

Fig. 1. (a) Schematic of the InGaN-based cyan LD grown on Si. (b) The InGaN active region with graded-compositional InGaN LWG layer, AlGaN cap layer and GaN/AlGaN/GaN multiple-layer QBs.

Download Full Size | PDF

The as-grown epitaxial wafer was processed into edge-emitting LD with a ridge width of 4 μm and a cavity length of 800 μm. Both rear and front facets were cleaved and coated with TiO₂/SiO₂ highly reflective dielectric multilayers to reduce the mirror loss and the threshold current. The surface morphologies of InGaN MQWs and LWG layers were obtained by atomic-force microscopy (AFM, Bruker Dimension Icon). The fine structures of InGaN active region were characterized by high-angle annular dark field (HAADF) scanning transmission electron microscopy (STEM). The on-bar electroluminescence (EL) spectra were performed by a fiber optic spectrometer (Idea Optics FX4000) under a pulse current injection with a pulse width of 400 ns and a repetition rate of 10 kHz. The output optical power as a function of the injected current was measured by an optical power meter (Thorlabs PM121D). The finite-difference time-domain method was applied to obtain the optical field profile of ridge-waveguide LDs.

3. Results and discussion

Graded-compositional InGaN LWG layer was utilized to enhance the optical confinement of InGaN active region, instead of conventional single-compositional InGaN LWG layer. Figure 2(a) is a diagram schematically showing the conduction band around the InGaN active region with graded-compositional InGaN LWG layer. The In composition was increased stepwise so that the band-gap energy decreased with the refractive index correspondingly increasing from the n-type AlGaN cladding layer towards the InGaN active region. Generally, an optical confining structure can be formed by using discontinuity refractive index at the semiconductor interface. Figure 2(b) presents the distribution of normalized 1D optical field intensity and the corresponding refractive index of InGaN-based cyan LD with single- and graded-compositional InGaN LWG layers. The In-content of single-compositional InGaN LWG layer was 3%, which was consistent with the average In-content of graded-compositional InGaN LWG layer. It can be observed that the InGaN-based cyan LD with graded-compositional InGaN LWG exhibits reduced optical field leakage outside of active region due to the enhanced optical confinement between the n-type AlGaN cladding layer and InGaN active region, compared to that of single-compositional InGaN LWG layer. With the refractive index increasing stepwise toward the InGaN active region, the barrier of the optical waveguide can be formed within the InGaN LWG layer (blue dashed line in Fig. 2(b)). The optical confinement factor of the InGaN active region was increased from 67% to 72% with this configuration, which allows for an improvement in the modal gain to initiate lasing.

 

Fig. 2. (a) Energy-band diagram of the InGaN active region with graded-compositional InGaN lower waveguide layer. (b) Profiles of the refractive index and optical field distributions of the InGaN-based cyan LDs.

Download Full Size | PDF

However, the growth of thick InGaN LWG layer typically results in a degraded surface morphology such as the increasing V-pits and surface roughing, which severely affects the quality of the following InGaN MQWs, as shown in Figs. 3(a) and 3(c) respectively. These V-pits would cause substantial non-uniform injection of carriers and significantly enhance the photon scattering at the heterointerfaces and thereby increasing the internal absorption loss of InGaN active region [20, 21]. Since the migration of Al adatoms is weaker than that of Ga adatoms and the Al-N bond energy is stronger than that of Ga-N, the Al adatoms that enter the V-pits are not easily desorbed, thus the AlGaN materials can be used to restore the V-pits and recovery a flat interface [23–25]. Figure 3(e) shows the dual-temperature growth scheme of the InGaN LWG layer and AlGaN cap layer. The growth temperature of AlGaN cap layer was 120 °C higher than that of InGaN LWG layer and low-temperature (LT) GaN cap layer. By introducing a 10-nm-thick Al_0.1Ga_0.9N cap layer, the density of V-pits was reduced from 4.8 × 10⁸ to 3.6 × 10⁷cm^-2 with the root-mean-square (RMS) surface roughness of InGaN LWG layer decreased from 0.3 to 0.1 nm, which exhibits a smooth step-flow morphology, as shown in Fig. 3(b). Meanwhile, the AlGaN interlayer and dual-temperature growth scheme were also adopted during the growth of high In-content InGaN MQWs to suppress In-content fluctuations and make the InGaN MQWs more robust, as shown in Figs. 3(d). Besides, a few amount of H₂ was introduced to eliminate the In-rich trench defects during the GaN/Al_0.07Ga_0.93N/GaN multiple-layer QBs growth after the deposition of InGaN QW cap layer [26]. As a result, the RMS surface roughness of the InGaN MQWs was decreased from 0.8 to 0.4 nm, and concomitant reduction in the density of V-pits. Moreover, a homogeneous emission pattern with no dark spots can be obtained in the micro-PL imaging of InGaN-based LD grown on Si (Supplement 1 Fig. S1), which indicates a high quality of InGaN-based green MQWs.

 

Fig. 3. AFM images of graded-compositional InGaN LWG layer and InGaN MQWs without (a), (c) and with (b), (d) AlGaN cap layer. (e) Schematic diagram of the growth process of InGaN LWG layer with Al_0.1Ga_0.9N interlayer.

Download Full Size | PDF

Figure 4 shows the cross-sectional HAADF-STEM images of the whole InGaN-based LD structure. It can be observed from Fig. 4(a) that a substantial portion of threading dislocations were filtered out by inserting a properly designed Al-composition step-graded AlN/AlGaN buffer layer. The enlarged image of active region is shown in Fig. 4(b). The thickness of InGaN QWs and QBs were determined as 2.5 nm and 11.0 nm, respectively, which was in line with the design. The GaN/AlGaN/GaN multiple-layer QBs consisted of an 1.0-nm-thick Al_0.07Ga_0.93N interlayer sandwiched between GaN QBs. An abrupt heterointerface between InGaN QW and QB along with a uniform thickness can be found in Fig. 4(b). This result further indicates that the AlGaN interlayer within the InGaN active region can improve the interface quality of InGaN-based green MQWs grown on Si. Moreover, no newly generated threading dislocations were observed in the InGaN green MQWs, which also benefits from both the graded-compositional InGaN LWG layer that reduces the in-plane lattice mismatch between InGaN MQWs and LWG layer and the strain compensation by introducing the AlGaN interlayer in QBs.

 

Fig. 4. (a) The cross-sectional HADDF-STEM image of the InGaN-based cyan LD grown on Si. (b) The enlarged image of marked rectangle in (a).

Download Full Size | PDF

The characteristics of the as-fabricated InGaN-based LD grown on Si were measured under pulsed current injection at room temperature. Figure 5(a) shows the electroluminescence (EL) spectra of the InGaN-based LD under pulse electrical pumping below and above the threshold current. The lasing wavelength was 480 nm with a threshold current of 585 mA, corresponding to a threshold current density of 18.3 kA/cm². The peak wavelength of EL spectra was blue-shifted from 510 nm to 480 nm when the injection current rose up from 200 to 650 mA due to the screening of quantum-confined stark effect. The spontaneous emission is dominant with a broad EL spectra at low injection current, but dramatically narrows down when the injection current is larger than 585 mA. Meanwhile, the far-field pattern (FFP) of the edge emission with a pulse injection current at 1.2 times threshold current was shown as the inset, exhibiting an elliptical pattern elongated along the growth direction due to the asymmetric optical confinement in the longitudinal and latitudinal directions [27]. The FFP and the narrowed EL spectra provide direct evidence of stimulated emission for LD. Figure 5(b) presents the light-current characteristic of InGaN-based cyan LD grown on Si. A clear keen point was observed at the injection current of 585 mA, beyond which the light output power increased much more rapidly.

 

Fig. 5. (a) EL spectra of the InGaN-based cyan LD grown on Si under pulse electrical pumping at room temperature. The inset shows the FFP obtained under the injection of 1.1 times of threshold current. (b) The light output power as a function of injection current.

Download Full Size | PDF

4. Conclusion

In conclusion, we have demonstrated a pulse operation of InGaN-based cyan LD grown on Si substrate at room temperature for the first time. The threshold current density is 18.3 kA/cm² and lasing wavelength for the cyan LD is 480 nm. By utilizing a 10-nm-thick Al_0.1Ga_0.9N interlayer to restore the V-pits and recovery InGaN heterointerfaces, the RMS surface roughness of InGaN LWG layer was significantly decreased and the density of V-pits was reduced one order of magnitude from 4.8 × 10⁸ to 3.6 × 10⁷cm^-2. Meanwhile, an enhanced optical confinement from 67% to 72% within the InGaN active region can also be found under the graded-compositional LWG configuration. Such a proper interface engineering of the InGaN LWG and MQWs finally leads to the demonstration of the InGaN-based cyan LDs grown on Si. The lasing wavelength was 480 nm with the threshold current of 585 mA, corresponding to a threshold current density of 18.3 kA/cm². This work paves the promising way for fabricating InGaN-based long-wavelength LDs grown on Si.