Advanced microscale silicon photonics technology has emerged as a promising candidate for the next-generation chip-scale data communication network due to its unique advantages of low cost, high integration density, high speed, and energy efficiency^[1,2]. However, a highly efficient microscale Si-based light source is still considered as the obstacle for realizing a practical Si-based photonic integrated circuit (PIC), due to the indirect bandgap nature of bulk Si and Ge materials.Various approaches towards integration of microcavity lasers on silicon substrates have been demonstrated extensively, including co-packing, hybrid integration, wafer bonding, etc. But, these methods more or less include cumbersome fabrication processes^[1,2]. The monolithic integration method, however, is another promising route towards low cost and scalable Si-based PICs. Monolithic integration takes advantage of the CMOS-compatible fabrication process of contemporary integrated circuits. It is unnecessary to consider heterogenous methods to make precise alignment between the light souce and other optical components. Nevertheless, the fundamental challenge of epitaxial III-V on Si is the degraded material quality caused by large material dissimilarities^[3]. The planar defect and anti-phase boundary, due to the polar on non-polar growth, can be avoided by employing a pretreated double-atomic-step Si substrate^[4]. Furthermore, a strained-layer superlattice (SLS) has been used as defect filter layers (DFLs) to significantly reduce the threading defect caused by the lattice mismatched heteroepitaxial growth of III–V materials on Si^[5]. Those works pave the way for high-performance ridge-waveguide lasers with quantum dot (QD) gain medium directly grown on the group-IV substrates through various intermediate buffer layers and offcut substrates, including Ge^[6], Ge-on-Si^[7], GaP/Si(001)^[8], patterned on-axis Si (001)^[9], and Si substrate with a 4° offcut angle^[3]. However, the relatively large footprint of the ridge-waveguide laser limits the realization of microscale Si-based PICs. Recently, we have demonstrated InAs/GaAs QD microdisk lasers and photonic crystal membrane lasers monolithically grown on CMOS-compatible on-axis Si (001) substrates^[10–12].