A model relating wearout induced physical changes in thin oxides to the statistical description of breakdown

Abstract

AbstnctA model has been developed that related the physical wearout caused during high voltage or high current stressing of thin oxides to the measured statistical TDDB distributions. Wearout was described in terms of changes that occurred in the oxide prior to breakdown.Traps were generated within the oxide during wearout and breakdown occurred locally when the local density of traps exceeded a critical value. The model has been used to quantify several effects observed during TDDB measurements. The area dependence of breakdown distributions, the differences in the breakdown distributions of constant current stresses and constant voltage stresses, and the multimodal distributions often observed were simulated using physically measured quantities and realistic assumptions about the trap distributions inside of the oxides. Several TDDB distributions measured by others were fitted using the model developed here. Introductioq

Oxide Wearout, Breakdown, and Reliability

“…This drop in trap generation rate is valid if either the stress fluence or stress time increases or if higher stress voltages are applied. This drop in the trap generation rate as the oxide wears out is consistent with the model of trap generation shown in Figure 21 and Figure 22 and was the first evidence that the bridging oxygen bonds were responsible for the traps in the oxide [448,477,490,621]. The drop in the trap generation rate is a measure of the dispersion in the bonding energy (angle) of the bridging oxygen bonds.…”

“…The trap generation rate was proportional to the fluence 173 , something universally measured for bulk oxide trap generation [351,494,495,524,543,621,634,684,690,691]. The trap generation rate falls as the stress progresses, supporting the model of the bridging oxygen bond being responsible for the traps [490,621,659,691]. The trap generation rate falls as the fluence" 273 , since this quantity is the derivative of the trap density.…”

“…This dependence is needed when using trap generation data and statistical models to calculate the TDDB distributions for different oxide thicknesses and fields [48,49,562,697,698]. The time dependence of trap generation has been reported by several workers [48,506,530,588,605,621,[700][701][702][703]. In general, the trap generation rate is fast initially and becomes slower as the density of traps increases, in agreement with a model of trap generation based on breaking the bridging oxygen bond, as described in Figures 21 and 22.…”

Oxide Wearout, Breakdown, and Reliability

“…The trap generation rate was proportional to the fluence" 3 , something universally measured for bulk oxide trap generation [351,494,495,524,543,621,634,684,690,691]. The trap generation rate falls as the stress progresses, supporting the model of the bridging oxygen bond being responsible for the traps [490,621,659,691]. The trap generation rate falls as the fluence" 2 ' 3 , since this quantity is the derivative of the trap density.…”

“…This dependence is needed when using trap generation data and statistical models to calculate the TDDB distributions for different oxide thicknesses and fields [48,49,562,697,698]. The time dependence of trap generation has been reported by several workers [48,506,530,588,605,621,[700][701][702][703]. In general, the trap generation rate is fast initially and becomes slower as the density of traps increases, in agreement with a model of trap generation based on breaking the bridging oxygen bond, as described in Figures 21 and 22.…”