Barrier heights and passivation of grain boundaries in polycrystalline silicon

Abstract

A simple frequently presented model for the electronic structure of a grain boundary in an n-type semiconductor is shown in Fig. 1. We have recently completed an extensive series of transport measurements on single grain boundaries in neutron-transmutat ion-doped polysilicon with the intent of investigating the applicability of this model. If thermal emission of electrons over the barrier predominates, the zero bias conductance Go across such a barrier will be proport ional to exp (---¢B/kT) and the observed activation energy Eac t of this quant i ty will be @S -TaC~B/a T. Seager and Castner [1] have observed such an Arrhenius behavior for Go above 270 K for large mismatch angle barriers in silicon; a plot of their activation energies versus dopant density is shown in Fig. 2. If CB is linear in temperature (as simple theory would suggest), Eac t is approximately CB(T = 0) and hence the vertical scale of Fig. 2 can be interpreted as the low temperature barrier height. The general Nd dependence of these data is in accord with expectations. As the donor density decreases an increasingly smaller amount of excess negative charge is transferred to the boundary (for a fixed value of @B). This leaves the Fermi energy in the barrier near its value in the neutral condition (EFB in Fig. 1). Thus the barrier height in this saturated regime should be constant and equal to EFG -EFB ; if EFB is temperature independent, this implies a constant value of Eact (as observed) equal to Ec -EFB.