Addressing the computational challenges of NEGF simulations

This thesis focuses on the challenges and advancements in nanoscale optoelectronic device simulation, with particular attention on the role of quantum mechanical effects on device performance. By incorporating the Non-equilibrium Green’s Function formalism, this work establishes a framework that combines quantum transport simulations, performed with the libNGEF library, with a multiscale TCAD software called TiberCAD. Due to the computational demands of the NEGF method, the research emphasizes efficiency, implementing optimized algorithms and parallelization techniques within libNEGF to facilitate realistic device modeling. A significant contribution is also the exploration of different Hamiltonian representations to balance accuracy and computational load. In this context, a novel Machine Learningbased correction scheme enhances the Empirical Tight-Binding method, improving its precision while maintaining efficiency. Furthermore, simulations incorporating quasi-equilibrium approximations show enhanced current-voltage and carrier density characteristics by self-consistently combining the quantum-computed density with a semi-classical Drift-Diffusion model. This approach, leveraging confined discrete states within quantum wells, allows the simulation of larger optoelectronic devices with reasonable accuracy. Finally, early-stage testing of new libNEGF features, such as scalability benchmarking and the inclusion of electron-phonon/electron-photon interactions, suggests promising avenues for future development, highlighting the framework’s potential to advance nanoscale device simulation.