对量子点激光器进展和与 Si 光子集成电路集成的展望

I nternet Protocol traffiffiffic has experienced a compound annual growth rate of 27% in the past few years, in which almost 75% of the data traffiffiffic resides within the data center. Leveraging well-established processing in Si-based microelectronics, Si photonics is expected to meet this soaring global demand for low power and high bandwidth density optical interconnects.Both Si and its native oxide are transparent in the commercially important datacom and telecom wavelength ranges and can form high-index contrast waveguides ideally suited for highly integrated photonic integrated circuits (PICs). A variety of high-performance passive components have been demonstrated on 300 mm Si wafers.8 Yet, due to the indirect bandgap of Si and Ge, realizing on-chip light sources requires integrating high quality III−V gain materials onto the existing Si photonic platforms. Compared to earlier hybrid integration where precise laserchip alignment is needed at the fifinal packaging stage, heterogeneous integration via wafer bonding has greatly simplifified the integration process and provides the prospect for scalable manufacturing. Through a decade of nonstop innovations in academia and industry, heterogeneous integration is currently reaching maturity, achieving mass commercial production lines. However, the cost and the size of III−V.

Monolithic integration through direct epitaxial growth is more economically favorable and provides a better heatdissipation capability. Yet, the performance of epitaxially grown quantum well (QW) devices on Si fall far behind native substrate devices due to the crystalline defects generated during the heteroepitaxial growth. QDs, fifirst introduced in 1982 by Arakawa and Sakaki, are important because the strong strain fifield induced by QDs hinders the in-plane movement of dislocations. Besides the defect insensitivity,QDs have numerous performance advantages over QWs, including lower threshold current, higher temperature operation, a near zero line width enhancement factor and, thus, isolator-free stability, ultrafast gain recovery, and enhanced four-wave mixing. Exciting technological advances have been made in the fifield of epitaxially grown QD devices on Si, as described in several previous review articles, summarizing the evolution of device performance, and efffforts in developing CMOS compatible epitaxial platform on (001) Si.

In this Review, we will present the most recent advances in monolithically integrated QD devices on a Si photonic platform, with a focus on breakthroughs in a long lifetime at elevated temperatures. After a brief introduction of the fundamental advantages of QDs, we will summarize several technological breakthroughs at the device and platform levels. Specififically, through the novel management of crystalline defects, the most recent breakthroughs in high temperature CW operation will be described. Finally, we end by providing a high-level summary of the difffferent generations of lasers and aging results and discussing the prospects in obtaining a monolithic integration in real-world applications.

With this much reduced diffffusion length in QDs, Fabry− Perot (FP) QD lasers grown on Si showed that reducing the ́ ridge width shows a continuous decrease of threshold currents down to a 2μm ridge width with no sign of threshold current density increase.Microring resonators fabricated from the same material demonstrate monotonically decreased threshold currents with the reduction of the ring radius. Threshold currents below 1 mA were achieved with a radii as small as 4 μm. Decreased carrier diffffusion length also reduces the sensitivity to crystalline defects. Since crystalline defects are unavoidable in epitaxially grown III−V material on Si, this makes QD emitters an ideal candidate for monolithic light source integration. Another important aspect that can be seen from Figure 1c,d is that the excited state transitions are more affffected by the nearby defects due to higher carrier escape rates. Thus, maintaining a ground state operation is essential for reliable and high performance epitaxially grown devices.

Fig1

Parallel to the continual improvement of material quality and individual QD device performance on Si by epitaxial growth, intense efffforts are also devoted to the development of fully integrated photonic circuits. Today, InPbased monolithic integration and Si photonics are the two major integration technologies in photonics. QD-based PICs on Si can thus be built upon what has been achieved through either monolithic or heterogeneous integration to both improved performance and reduced cost.

To achieve low-loss active− passive coupling of III−V with Si waveguides, heterogeneous integration offffers an elegant path, as shown in Figure 5a. Ever since the invention of the fifirst prototype QW-based heterogeneous laser in 2006,51 it only took 10 years for Intel to announce the fifirst commercial products. The successful commercialization of InP-based QW-based heterogeneous integration indicates that GaAs-based QDs for 1300 nm and InAs/InP QDs/quantum dashes for 1550 nm should be straightforward engineering, and some of the lessons can be leveraged, for example, effiffifficient bonding strength, reliable operation, adoption by the CMOS production foundries, and effiffifficient active−passive optical coupling. Changing from InPbased QW epi to GaAs-based QD epi requires difffferent material systems and, thus, completely difffferent device design, as well as process optimization. While research has been conducted actively in both heterogeneous QW lasers and QD lasers, combining the two together is still at an embryonic stage.