Nano Semiconductor Devices

As the size of the semiconductor shrinks to nano meters, the quantum effects in the device become important. This makes it necessary to simulate the device using quantum mechanics, instead of semiclassical methods. While most such quantum mechanical device simulations use continuous effective mass theory, here we have used atomistic method to simulate the device. There are many advantage of using atomistic methods. For example, it can describe the single dopant effects more naturally, and it can also describe surface fluctuations. Besides, for the size regime of a few nanometers, the effective mass method might become inadequate.

We are using empirical pseudopotential method to desribe the Hamiltonian of the system, while using linear combination of bulk band (LCBB) method to calculate the eigen states of the system. LCBB enables us to calculate million atom systems on a single CPU in a few hours. For device with quantum transport, we have developed a method, which decomposes the single electron eigen states into left and right injecting scattering states. This method yields very accurate results compared with direct scattering state calculations, but with the eigen state problem cost (the LCBB cost). Using this method, and the Poisson solver (e.g., FlexPDE) to have a selfconsistent calculation for the carrier density, we plan to study different devices for their I/V characterizations and other properties.


A few examples are given here for the systems studied in this group.

The electron and hole wave functions in a pyramidal InAs quantum dot embedded in GaAs matrix calculated using LCBB method. .


The carrier charge density near the gate at the turn-on threshold gate potential. This is part of the simulation for a CMOS device.


The I/V curve of a double gate FET. Our calculation TBS is compared with the standard Non equilibrium Green's function (NEGF) method based on tight-binding Hamiltonian. We found well agreement between these two methods. But TBS is run on a single processor machine, and the NEGF is run on a supercomputer.