DIFFERENT METHODS OF NOISE-REDUCTION IN MONTE-CARLO SIMULATIONS OF A SCHOTTKY DIODE

Methods of decreasing the noise and increasing the computational effic iency in Monte Carlo simulations of semiconductor devices are investig ated. A lookup table approach to the charge assignment to mesh points has been implemented which is independent of the complexity of the ass ignment function. Using this approach reduces the CPU time of the char ge assignment to about one third. Potential fluctuations in low held r egions are compared for different assignment function, which shows the advantage of using more complex schemes than those generally used. Re sults from a full band Monte Carlo simulation of a submicron Schottky diode are presented and the values of the current density for differen t assignment schemes are compared, showing a lower noise for the highe r order schemes. Statistical enhancements by splitting of superparticl es in the depletion region has been investigated. The combination of a high order assignment scheme and statistical enhancement by splitting decreases simulation runtimes considerable for a given noise toleranc e.