An efficient molecular dynamics scheme for predicting dopant implant profiles in semiconductors

Abstract

We present a highly efficient molecular dynamics scheme for calculating the concentration profile of dopants implanted in group-IV alloy, and III–V zinc blende structure materials. Our program incorporates methods for reducing computational overhead, plus a rare event algorithm to give statistical accuracy over several orders of magnitude change in the dopant concentration. The code uses a molecular dynamics (MD) model, instead of the binary collision approximation (BCA) used in implant simulators such as TRIM and Marlowe, to describe ion-target interactions. Atomic interactions are described by a combination of `many-body' and screened Coulomb potentials. Inelastic energy loss is accounted for using a Firsov model, and electronic stopping is described by a Brandt–Kitagawa model which contains the single adjustable parameter for the entire scheme. Thus, the program is easily extensible to new ion-target combinations with the minimum of tuning, and is predictive over a wide range of implant energies and angles. The scheme is especially suited for calculating profiles due to low energy, large angle implants, and for situations where a predictive capability is required with the minimum of experimental validation. We give examples of using our code to calculate concentration profiles and 2D `point response' profiles of dopants in crystalline silicon, silicon–germanium blends, and gallium–arsenide. We can predict the experimental profile over five orders of magnitude for 〈100〉 and 〈110〉 channeling and for non-channeling implants at energies up to hundreds of keV.