Stress Induced Increased Low Level Leakage in Thin Oxides

Abstract

It has been observed that the low-level, pre-tunneling currents through thin gate oxides increased after the oxides had been stressed at high voltages. The number of traps inside of the oxide generated by the stress has been shown to increase as the 1/3 power of the fluence that had passed through the oxide during the stress. The increases in the low-level, pre-tunneling currents have been shown to be proportional to the number of stress generated traps in the oxide and not to the fluence during the stress. The voltage dependences of the excess low-level leakage currents were stress and measurement polarity dependent. Attempts have been made to fit the voltage dependences of the excess low-level currents to Fowler-Nordheim tunneling, Frenkel-Poole conduction or Schottky barrier lowering. The increase in the portion of the low-level, pre-tunneling current that was not dependent on stress/measurement polarity sequence was best fit using Schottky emission currents. The model that has been developed to describe the increases in the low-level currents has centered on trap-assisted currents through the oxides.

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

“…and to trapassisted tunneling [ 151. The voltage dependence of these low-level currents was complex [7].The currents were higher when the stress polarity and the J-V polarity were opposite than when they were the same.…”

“…It should be noted that the voltage dependence of the low-level current that flowed through the oxide after stress was a function of the stress fluence, the polarity of the J-V measurement voltage compared with the polarity of the stress, the amount of time that had elapsed between the stress and the J-V measurement, and whether a previous J-V measurement had been made 161. The portion of the low-level pre-tunneling current that was independent of time of measurement and measurement polarity sequence had a voltage dependence expected from a Schottky emission current [7]. The current that flowed through an 8nm thick oxide fabricated on n-type silicon during and after voltage pulses had been applied to the oxide has been shown in Figure 3, with the magnitude of the applied voltage being the variable parameter.…”

Oxide Wearout, Breakdown, and Reliability

“…The RILC is probably due to the removal of the mobile carriers generated by the irradiation and not by trap generation inside the oxide. When the RILC was measured, not by sweeping the voltage but by keeping the voltage fixed in the pre-tunneling regime and measuring the current as a function of time, the current was found to decay as t" 1 , similar to the t" 1 decay of SILCs [325,327]. This t" 1 component of the RILC decreased as the RILC decayed with storage time of the capacitor, similar to effects measured on other irradiated oxides [39].…”

“…The dc component dominates at higher measurement voltages and the transient component dominates at lower measurement voltages [355]. Both components are proportional to the density of the stress-generated traps [325,355]. The transient component has been shown to be due to tunnel charge/discharge of the stress generated traps and is proportional to f' [327].…”

Oxide Wearout, Breakdown, and Reliability

“…The SILCs are stable at room temperature, with no reduction in the SILC being measured one year after the stress was removed [325]. It is important to take care when measuring SILCs, since there is a one-time-only component to the SILC that is measured whenever the measurement polarity is changed.…”