The effect of enduring charge injection on the physical properties of the SiO2 layer of a metal-oxide-semiconductor structure is studied by means of a novel characterization method. It is based on the observation reported previously, that under charge injection conditions the density of occupied oxide traps reaches a value which is only a fraction of the total trap density. This trap occupation level is strongly dependent on the oxide electric field. The oxide trap density can be evaluated by measuring this field dependence, using a relatively small amount of charge injection. This method is used to distinguish between the process of trap generation and electron trapping in the generated traps, under conditions of continuous charge injection up to levels of more than 50 C/cm2. The trap generation rate is found to be proportional to the flux of the injected charge, and to increase exponentially with the oxide electric field. At high oxide field only a small fraction of the newly generated traps are occupied; consequently, the measured oxide charge buildup does not reflect the actual increase in the density of generated traps. The density of the generated traps reaches high values of the order of 1020 cm−3. It is suggested that these high values of oxide traps may be the cause of the SiO2 ‘‘wear out’’ type breakdown, by forming a new path of conductance by electron tunneling between closely spaced generated traps.

No abstract

The field and time dependence of charge carriers trapping under different charge injection conditions, is studied in this work, using the dc hot electron injection technique. It is shown that the trapping characteristics converge to field-dependent quasisaturation values. Variation of the trapping levels, due to change of the oxide field magnitude, are obtained in both directions and exhibit complete reversibility. These results, which cannot be explained by the first-order conventional trapping model, are consistent with a dynamic trapping-detrapping model. According to this model, quasisaturation of trapping characteristics is obtained when the trapping and detrapping processes are balanced. The occupation of the traps under steady-state conditions is therefore field dependent. The same model also describes the generation of positive charge under high-field injection conditions. This phenomenon is shown to be related to ionization of localized states in the SiO2 forbidden gap. The implications of the dynamic model suggest the need for a reevaluation of the present characterization methods for trapping in oxide layers.

The generation of a bulk positive charge in SiO2 layers of silicon gate metal-oxide-silicon (MOS) devices, under the conditions of high-field and charge injection is studied. The time dependence of the positive charge and its spatial distribution as a function of the oxide thickness and electric field are all consistent with an impact ionization-recombination model which takes into account both the spatial and the field dependence of the ionization probability. The nature of the ionization, either band-to-band, or traps ionization, is still unknown. Bulk positive charge of the same nature is also formed in Al gate oxides. Nevertheless, it was not always observed in previous works since a much larger Si-SiO2 interfacial positive charge is also generated in these samples.

The formation of bulk positive charge in SiO2 induced by charge injection and high field conditions is studied in this work, utilizing new experimental methods. Using a dc hot electron injection technique it is shown that the presence of electrons in the SiO2 conduction band is a necessary condition for the positive charge formation. This result rules out field (only) induced mechanisms as possible explanations of high field positive charge generation in SiO2. The positive charge formation and annihilation are found to be governed by the same rate equation, and, therefore, exhibit similar behavior as a function of time. This behavior is consistent with an impact ionization-recombination model, with a recombination cross section σ=2×10−16 cm2 and ionization rate α which is in agreement with previous published values. However, other collision activated mechanisms cannot be ruled out.