High-Reliability InAs QD Lasers on Silicon Through Misfit Dislocation Trapping Layers

Abstract

Abstract Body: Silicon photonics require a reliable integrated light source to be a viable technology. While heterogeneous integration techniques such as wafer bonding III-V on silicon has proven successful, dramatic cost reductions can be achieved through direct growth of III-V on silicon. This presents numerous challenges to material quality due to mismatches in lattice constant, coefficient of thermal expansion, and crystal symmetry. Even so, decades of work developing defect filtering buffers paired with defect-tolerant quantum dot (QD) active regions have yielded lasers with satisfactory reliability at room temperature, however laser lifetimes at elevated temperatures remain insufficient for commercial applications. We have recently uncovered a source for lagging performance in InAs QD lasers on silicon, namely misfit dislocations (MDs) that lie along the upper and lower boundaries of the active region and act as potent non-radiative recombination centers. We have proposed that these MDs form not due to lattice mismatch in the active region, but instead form by an alternative mechanism based on (1) tensile thermal strain that builds in the AlGaAs cladding layers during cooldown after growth and (2) pinning of threading dislocations (TDs) in the InAs QD active region. TDs, which are mobile in the cladding layers above ~300 °C, glide to relieve the thermal tensile strain in these layers. The TD segment passing through the active region would ideally glide with the segment in the cladding, but as QDs act as powerful pinning sites, glide is prohibited under these low stress conditions. It follows then that MDs will form at both the upper and lower boundaries between the cladding and active region if both cladding layers are above critical thickness due to thermal strain. To solve this, we insert strained indium-alloyed “trapping layers” (TLs) to displace formation of these MDs away from the active region. Unlike the active region, these layers do not contain QDs, but they can still pin TDs in the low stress environment encountered during cooldown because natural compositional fluctuations in the indium-alloyed TLs generate sufficiently large in-layer stress fluctuations. In lasers with TLs, MDs are no longer adjacent to the active region, so their supply of minority carriers is sharply reduced, benefiting not only initial laser performance but also laser reliability since this reduces recombination-enhanced dislocation climb, which we have observed to be a significant source of laser degradation. Initial performance improvements over baseline include median threshold current reduced by half, median slope efficiency increased by over 50%, and median output power increased by over 3×. Thermal performance is also greatly improved with the threshold characteristic temperature, T0, increased from 70 K in baseline lasers to 100 K in TL lasers, indicating their lasing threshold is less sensitive to temperature. Initial laser aging experiments reveal that TLs dramatically enhance the reliability of InAs QD lasers on silicon. After aging for 1200 h at 60 °C at approximately 2× initial 60 °C threshold current, TL lasers have a median threshold current increase of just 3%, far lower than the 56% increase for baseline lasers, lower than the 19% increase of our previous record lifetime InAs QD lasers on silicon under comparable conditions, and even lower than the 11% median threshold increase of QD lasers on GaAs substrates. We suspect that the TL lasers’ lower degradation rate compared to lasers on GaAs, despite their much lower TD density, is due to the higher thermal conductivity of the silicon substrate which reduces the active region temperature during operation. Overall, these reliability results provide a clear path forward toward reliable monolithically integrated light sources for silicon photonics.