One-dimensional silicon nanostructures, such as silicon nanowires, have diverse potential applications in field-effect transistors, thermoelectric devices, biological and chemical sensors etc. However, its thermal properties and heat dissipation have significant influence on the function and performance of these devices, which has not been investigated comprehensively in physical experiments.
By developing a scalable and highly efficient nonequilibrium molecular dynamics (NEMD) simulation method base on bond-order potentials, we are now able to overcome the long-lasting finite size effects in the computation of the thermal conductivity of silicon nanowires and obtain a thorough knowledge of its mechanical and thermal properties using the Sunway supercomputer. The longitudinal dimensions of the simulations exceed far beyond microscale, and more importantly, the lateral characteristic sizes are much larger than 10 nanometers, explicitly comparable with the silicon nanowires fabricated and measured experimentally, whereas the traditional simulation size is of several nanometers. Our largest simulations made use of 131,072 deeply fused many-core processors, namely, about 8.39 million computing processing units, where the parallel efficiency were above 80% and the computing performance on a single many-core processor reached 15.1% of the theoretical peak performance.
This virtual experimental measurement implemented on the Sunway supercomputer also paves an exciting and promising way to deeply explore the thermophysical properties of other low-dimensional structures of covalent solids, such as carbon nanotubes, grapheme and silicon nanosheets.