Intrinsic limitations on ultimate device performance and reliability at (i) semiconductor–dielectric interfaces and (ii) internal interfaces in stacked dielectrics

Abstract

The scaling of electrical oxide thickness to 1.0 nm and below for advanced silicon devices requires a change from thermally grown oxides and nitrided oxides to deposited dielectrics which have dielectric constants, k, significantly greater than that of silicon dioxide, k0∼3.8. Implementation of the higher-k dielectrics into field effect transistor devices requires a processing protocol that provides separate and independent control over the properties of the Si–dielectric interface and the bulk dielectric film. Experiments to date have shown that plasma-grown nitrided oxides, ∼0.5–0.6 nm thick, satisfy this requirement. This paper addresses chemical bonding issues at the Si–dielectric interface and at the internal dielectric interface between the plasma-grown nitrided oxides and the high-k alternative dielectrics by applying constraint theory. Si–SiO2 is a prototypical interface between a “rigid” Si substrate and a “floppy” network dielectric, SiO2, and the interfacial properties are modified by a monolayer-scale transition region with excess suboxide bonding over what is required for an ideal interface. Additionally, the defect properties at the internal interface between a nitrided SiO2 interface layer and a bulk dielectric film reflect differences in the average number of bonds/atom, Nav, of the dielectrics on either side of that interface. Experimentally determined interfacial defect concentrations are shown to scale quadratically with increasing differences in Nav thereby establishing a fundamental basis for limitations on device performance and reliability.