Negative Bias Temperature Instability (NBTI) is one of the most pressing reliability issues in modern CMOS semiconductors. It is caused by application of a negative gate voltage to a MOSFET and manifests itself as a shift in the device parameters, first and foremost the threshold voltage. The effect is considerably enhanced at elevated temperatures, and p-channel FETs are more affected than n-channel FETs. Although a great deal of effort is going into the research on this topic, there is still no universally accepted microscopic model.
As with modeling in general, accurate and reliable measurements are the key to successful model development. Especially with sophisticated devices in the sub-micron range, second-order effects cause the measured device characteristics to deviate considerably from the idealized closed-form analytical solutions derived from the semiconductor equations. It is therefore important that measurement data be assessed by means of simulation of the device within the chosen measurement setup. This is true to an even greater extent in the case of NBTI, since measurement data from different groups often do not match and sometimes even contradict each other. This discrepancy is largely due to the fact that the relaxation of NBTI sets in very quickly, but lasts for several decades in time.
One particular measurement method called On-The-Fly (OTF) avoids relaxation by continuously stressing the device while measuring the drain current degradation. Since no degradation is 'lost' during some measurement cycles, this method was considered superior in the past. Recently, the concern was raised regarding the extent to which errors in the initial measurement of the drain current, to which all subsequent drain current readouts are compared to, may impair the calculated degradation of the threshold voltage. Our simulations clearly showed that uncertainties in the initial drain current measurement seriously impact the extracted threshold voltage shift. In addition, mobility variations induced by NBTI may add spurious threshold voltage shifts.