Full-Band Monte Carlo Transport Calculation in an Integrated Simulation Platform

Integrated Simulation Platform U. Krumbein D. Yoder A. Benvenuti A. Schenk W. Fichtner Integrated Systems Laboratory, ETH{Z urich, Switzerland Phone: +41 1 632 65 44, FAX: +41 1 632 11 94 E-mail: [email protected], [email protected], [email protected], [email protected], [email protected] The Monte Carlo method of charge transport simulation o ers the possibility to extract information about all quantities derivable from the semiclassical distribution function, whose accuracy is limited explicitly by statistical convergence and implicitly by the quality of the physical models. To date, much e ort has been devoted to improving models for bandstructure and scattering mechanisms, such as electron-phonon scattering, impact ionization and other carrier-carrier scattering [1]. However, the practical usefulness of Monte Carlo device simulation has not entirely lived up to its promise, as evidenced by the observed propensity to simulated simplied or imaginary device structures. A unique device simulation environment has been developed which unites the capabilities of process, drift-di usion/hydro, and Monte Carlo simulation into a single platform. One may use Dios ISE [2] to begin with a process simulation. Drift-di usion or hydrodynamic simulations can be performed with the mixed-modemulti-dimensional device simulator Dessis ISE [3, 4] as a preprocessing step. The full-band Monte Carlo simulator Vegas ISE [1] was imbedded into Dessis ISE by a window technique. The domain of the Monte Carlo simulation may be chosen either as the entire device, or as a simply connected subregion. When the Monte Carlo simulation is invoked, it uses the precise device structure which has been generated by the process simulation. Former implementations of the window technique used the drift-di usion information as initial and boundary conditions [5, 6]. In this paper initial and boundary conditions are extracted from a hydrodynamic solution calculated by Dessis ISE. Carrier densities, velocities and temperatures are passed to the Monte Carlo part. The Monte Carlo simulations can be performed self-consistently or using a frozen eld from Dessis ISE, either in one or two dimensions. doping.col.ps 49 28 mm Source Gate Figure 1: 0.5 m nMOSFET with arbitrarily shaped Si-SiO2 interface from process simulation. We present three examples: a 0.5 m MOSFET, a 40nm MOSFET and a 0.5 m nin structure. The 0.5 m nMOSFET was fabricated and measured by Fujitsu. The process was simulated with Dios ISE leading to a non at SiSiO2 interface (Figure 1). Figure 2 shows that the drift-di usion Ansatz completely fails in this example. sisdep.drain.epsi 55 41 mm