methodology

Methodology

Simulation Framework

As the device size scales down, the focus shifts from transport in the Channel region to (Hetero)Junction, Interface, Contact, and Surface. Number of possible combinations of interfaces and junctions grow very fast: More and more elements in the periodic table and their combinations are considered for channel and dielectric materials (e.g. Si, Ge, III-V, 2D, high-k/low-k dielectrics). To cope with this, we not only employ the empirical tight-binding or k.p method for the device Hamiltonian but also the first-principles based Hamiltonian.

• Poisson-NEGF Self-consistent calculations

Device simulations are performed by self-consistently solving the transport equation and electrostatics. As for the transport, we mainly deal with the quantum mechanical transport of charges governed by the Schrödinger equation, which is solved by the state-of-the-art non-equilibrium Green’s function (NEGF) method. The electrostatics given by the Poisson equations is solved by either the finite difference or finite elementmethods.

• All-electron NEGF calculations

For special devices like Tunneling FET, ordinary excess charge model cannot give correct solution. Their results are affected by selection of negative/positive carrier contribution. To solve this problem, we use all-electron calculation for NEGF formalism. Instead of excess carrier, we directly calculate electron and core charge in given non-equilibrium configuration. By separation of equilibrium (and core charge contribution) and non-equilibrium charge correction term, this calculation can be efficiently done using contour integral. With this algorithm, we can correctly handle complicate subjects like TFET, MS junction.

• Electron-phonon scattering

In order to assess the true potential of the nanoscale devices, it is essential to capture the phonon scattering effects together with the quantum effects such as quantum confinement and tunneling. The electron-phonon interaction has been treated in the framework of the nonequilibrium Green’s function method through the self-consistent Born approximation. We have developed an efficient scheme to calculate the electron-phonon scattering self energy in the mode space which greatly reduce the computational cost of SCBA.

Machine Learning

As the transistor scales down, the number of design variables which significantly alter the performance of devices exponentially increases, and more complex models are required to capture the effects of the variables. It is necessary to handle the complexity of TCAD modeling by utilizing the machine learning (ML)-based approach. In this work, we have constructed the framework for the ML-based device optimization with TCAD. TCAD calculates the device performance of the design given from ML and gives it to ML. ML recommends the next calculation candidate in the direction of optimum. The best design can be found quickly by repeating this automatic feedback process.