Electronic Structure Theory of Radiation-Induced Defects in Si/SiO2

Radiation-induced defects in metal-oxide semiconductor (MOS) materials are a major problem for MOS devices exposed to high-energy radiation such as in space-based applications. Charge trapping at these defect sites degrades the current/voltage performance of MOS devices and ultimately leads to device failure. A physical understanding of the structure and formation mechanisms of these defects is becoming increasingly important as the size of these devices continues to decrease. Because these defects are highly localized, ab initio quantum chemical methods applied to model clusters have been very successful in providing insight into their local structure as well as physical properties such as electron spin resonance (ESR) spectra. They also give insight into the formation mechanisms of these centers. Of interest to us is the identification and characterization in amorphous silicon dioxide (a-SiO2) of what are collectively known as E' centers. These centers all appear to be characterized by an unpaired electron that is strongly localized on the dangling sp orbital of a Si atom that is bonded to three O atoms. We report ab initio Hartree-Fock calculations that we have performed on possible models for oxygen vacancy sites: a 5-Si center, a 4-Si center, and a 2-Si center model. Clusters of increasing size and flexibility are used to verify that the properties that are computed are insensitive to the sizes of the clusters that are used. They are also used to demonstrate the effects of SiO2 network flexibility on the mechanism of defect formation. The clusters employed in these studies ranged from 28 to 87 atoms in size. These calculations were performed using STO-3G and DZP (Double zeta+polarization) basis sets and are only made possible by the availability of high performance computing resources. Our calculations confirm the previous work of Chavez, et al. that the E'δ center, characterized by a nearly isotropic 100G hyperfine splitting, is a positively-charged symmetrically-relaxed 2-Si center in which the unpaired electron is equally shared by the two Si atoms. Another center, characterized by a 420G hyperfine splitting and known as the E'γ center, is believed to be a positively-charged and asymmetrically-relaxed 2-Si center in which the unpaired electron is localized on one of the Si atoms and the positive charge on the other Si. This model, in which the positively charged Si atom is stabilized by forming a bond with an O atom (which becomes 3-fold coordinated) from the surrounding SiO2 network, was first proposed by Feigl, Fowler, and Yip for a similar defect known as the E1 center in α-quartz. Formation of E'δ or E'γ centers by hole trapping at 2-Si vacancy centers is found to be determined by the flexibility of the local SiO2 network. This stands in contrast to the prior work using semi-empirical and DFT methods that have suggested that the symmetrically relaxed and asymmetrically relaxed configurations are bistable minima. No evidence of bistability is found in our calculations.