Session: 08-01: Micro/Nanoscale Heat Conduction

Paper Number: 122172

122172 - Extreme-Scale Simulation of Heat Conduction of Silicon-Based Nanostructures and Devices 

Abstract: 

Low-dimensional silicon nanostructures are widely applied in such devices as field effect transistors, novel photovoltaic cells, thermoelectric arrays, integrated circuits and chemical sensors, where their thermodynamic properties and behaviors have a significant impact on the performance and stability of these devices. Molecular dynamics simulation is a powerful tool for studying the thermodynamic behaviors of silicon nanostructures and devices. However, the computational workload of the complex many-body interactive potential between atoms used in the simulation is heavy, accordingly, the calculated feature size of nanostructures is much smaller than the macroscopic experimental scale. And it is difficult to overcome the finite size effect of thermal conductivity calculation in non-equilibrium molecular dynamics simulation (NEMD). In addition, for some physical circuits consisting of a large number of nano-transistor devices, the macroscopic continuous methods are hard to treat such nanoscale factors as doping, thin dielectric layer, surface and interface in the device, while the microscopic quantum mechanics methods can only calculate one or several nano-transistors.

Herein, a highly efficient parallel molecular dynamics simulation framework is established for the large-scale simulation of silicon nanostructures and devices. In terms of one-dimensional nanowires and two-dimensional nanofilms, the simulated feature sizes have reached the range of experimental preparation and measurement, and the size in heat transfer direction exceeds millimeters. The variation of thermal conductivity and phonon transfer with the feature sizes of silicon nanostructures is illustrated based on the surmounted size effect of the NEMD method. For the simulation of massively integrated devices, the effects of the surface, interface, doping, dielectric layer, gate, and layout are involved, calculating the heat conduction of two different functional modules of physical circuits with approximately 800,000 nano-transistors and 1.31 trillion atoms. The simulations used up to 40 million processor cores for the massively parallel computation, which enable the virtual experiments on real-world nanostructured materials and devices for predicting macro-scale thermodynamic properties and behaviors from micro-scale structures directly. The simulation capability will bring about many exciting new possibilities in nanotechnology, information technology, electronics and renewable energies, etc.

      Furthermore, a series of corrected models for phonon relaxation times were presented and implemented to solve the mesoscopic phonon Boltzmann transfer equation for nanostructures and devices. A new Lattice Boltzmann Method (LBM) scheme with effective correction of phonon mean free path (MFP) and relaxation time is proposed to describe phonon heat transfer in silicon thin films and silicon medium with defects, where the correction factor is dependent on the lattice structure of LBM. And the thermal conductivity of silicon medium with defects decreases exponentially as the defect density increases. Compared to the literature, the accuracy of thermal conductivity calculated by the models was significantly improved. The above work effectively advances the accurate simulation of heat transfer processes in silicon-based nanostructures and devices.

Presenting Author: Chaofeng Hou Institute of Process Engineering, Chinese Academy of Sciences

Presenting Author Biography: Dr. Chaofeng Hou is a professor at Institute of Process Engineering (IPE), Chinese Academy of Sciences (CAS). His research interests are focused on multiscale electrothermal simulation of nanostructured materials and nanodevices, high-performance algorithm design for atomistic simulation, mesoscience theories for nanoscale heat transfer and gas-liquid-solid phase transition. He is also the author of nearly 40 research papers.

Authors:

Chaofeng Hou Institute of Process Engineering, Chinese Academy of Sciences
Aiqi Zhu Institute of Process Engineering, Chinese Academy of Sciences
Yufeng Huang Institute of Process Engineering, Chinese Academy of Sciences