Growing III-V semiconductor lasers directly on silicon circuitry will transform information networks. Currently, dislocations limit performance and lifetime even in defect tolerant InAs quantum-dot (QD)-based devices. Although the QD layers are below the critical thickness for strain relaxation, they still contain long, previously unexplained misfit dislocations which lead to significant non-radiative recombination. This work offers a mechanism for their formation, demonstrating that the combined effects of thermal-expansion mismatch between the III-V layers and silicon and precipitate and alloy hardening effects in the active region generate the misfit dislocations during sample cooldown following growth. These same hardening effects can be leveraged to mitigate the very problem they create. The addition of thin, strained, indium-alloyed trapping layers displaces 95% of the misfit dislocations from the QD layer, in model structures. In full lasers, performance benefits from adding trapping layer now both above and below the QD layers include a twofold reduction in lasing threshold currents and a threefold increase in output powers. These improved structures may finally lead to fully integrated, commercially viable siliconbased photonic integrated circuits.
MAINSilicon-based photonic integrated circuits will dramatically increase data network bandwidth and energy efficiency and enable new paradigms in chip-scale sensing, detection, and ranging. Direct crystal-growth methods integrating III-V semiconductor lasers with silicon promise cost-effectiveness and scalability, [1] however, fabricating reliable, highperformance GaAs-or InP-based lasers on silicon has proven challenging. [2][3][4] Lattice constant mismatch between the silicon substrate and III-V film generates dislocation line defects, including 'threading' dislocations, which rise upward through the film. [4] Where they intersect the device's active region, they facilitate non-radiative recombination, degrading performance. The energy released causes dislocations to lengthen during device operation, a run-away degradation process ending in device failure. [2,[5][6][7] Despite decades of work to reduce threading dislocation densities to 10 6 -10 7 cm -2 (refs [8-10]) [8][9][10] and to develop dislocation-tolerant active materials such as InAsquantum dots (QDs) in quantum wells (QW) (dots in a well or DWELL), [11][12][13][14][15][16][17] threading dislocations continue to stifle the development of commercially viable III-V lasers on Si. [7,18,19] We have recently identified the root of this contradiction using plan-view scanning transmission electron microscopy (PV-STEM): unexpected dislocations found lying flat along the uppermost and lowermost QD layers, in even record lifetime QD lasers. [20,21]