A nonlinear multigrid method for one-dimensional semiconductor device simulation: Results for the diode (Q911239)

The author studies a multigrid method for the solution of the 1-D problem (1-D diode): \(div(\epsilon \text{grad} \psi)=qn_ i(\bar p-\bar n+D)\) and \(div(\mu_ n(\text{grad} \bar n-\bar n \text{grad}(\alpha \psi)))=0,\) \(div(\mu_ p(\text{grad} \bar p+p \text{grad}(\alpha \psi)))=0,\) where \(\bar n=n/n_ i\), \(\bar p=p/n_ i\), \(\bar D=D/n_ i\); \(\psi\), n, p are the electric potential, the electron and hole densities, respectively; \(\epsilon\), q, \(\alpha\) are constant values; D is a given (nonsmooth) function of x; \(\mu_ n\), \(\mu_ p\) are functions of x, \(\psi\), n, p; \(n_ i\) is a wild function of x. Boundary conditions are of Dirichlet type or of Neumann type. A convergence parameter that is independent of the meshwidth is found; the reduction of the iteration error is reached in few iteration steps.