Stressing and high field transport studies on device-grade SiO2 by ballistic electron emission spectroscopy

The tip of a scanning tunneling microscope was used to inject hot electrons across the gate and into the oxide of a metal-oxide-semiconductor structure. This method, known as ballistic electron emission microscopy (BEEM), allows an arbitrary choice of the energy of the injected electrons, which may be further accelerated by the application of a gate bias. The high current densities and choice of energy make BEEM an attractive method to study hot electron transport and breakdown phenomena in dielectrics. The studies reported here were made on Pd/SiO2/Si(100) structures with a SiO2 layer thickness of 3.8 nm. Monte Carlo techniques were used to calculate the spreading of the electron beam as it traverses the oxide. A strong dependence of the spreading on the kinetic energy and oxide thickness were observed. Using the calculated beam spreads to determine current densities and injected charge densities, the charge-to-breakdown (Q(bd)) was measured for several breakdown sequences. The Q(bd)'s consistently exceeded by several orders of magnitude the values obtained by conventional Fowler-Nordheim (FN) tunnel injection under high field conditions. Most of the time, breakdowns could not be achieved for 3.8 nm oxides. It is concluded that impurity/defects still control all observed breakdowns; an intrinsic limit-although claimed to have been reached-has not yet been established. Changes in the BEEM spectra with injected electron charge are interpreted in terms of a three stage process to breakdown: (I) electron trap creation and filling at the SiO2-Si interface, (II) prebreakdown believed to occur through thinning of the oxide that starts at the oxide-Si interface, and (III) oxide punch-through, characterized by an injection threshold less than or similar to 1 V, close to that for Si. The role of hot hole injection into the SiO2 valence band was also assessed and deemed a negligible factor in the degradation process under the zero or low oxide biases used in the experiment reported here. (C) 1996 American Vacuum Society.