Analysis of the OFIT Technique

5.5 Analysis of the OFIT Technique

As described in [95, 96], a constant base-level CP measurement with $V = 2V Base$ is performed using a gradually increasing pulse amplitude $ΔV G$ . Until the desired stress level is reached, starting from $− 1V$ down to $− 18V$ , the pulse slopes have to be kept constant to obtain comparable results. Constant pulse slopes ensure that the upper and lower energy boundaries of the active energy interval remain unchanged when $ΔVG$ increases [50]. Due to a constant pulse slope the amplitude of $ΔVG$ is proportional to the pulse rising (often referred to as leading) and falling (trailing) time. Given the additional requirement of a constant duty cycle, the rise and fall times have to be adapted at every voltage step within the CP measurement to obtain the proper charge pumping current $Icp$ . Since it is inevitable to change both the pulse width and also the rise and fall times one has to ask for the potential pitfalls: Are OFIT-data obtained during stress and relaxation comparable? If that is not the case, is there some possibility to correct this nonconformity? These questions will be examined in the following.

Starting with Fig. 5.13 the two large arrows pointing up and down reveal some important aspects of the temporal evolution of the pulses during a CP measurement. The charge pumping current $Icp$ at stress conditions ( $VG,low < − 3V$ ) differs a lot when compared to that obtained during relaxation ( $VG,low = − 1V$ ). The higher the NBTI stress conditions, the larger the $Icp$ -signal becomes. This can be partly attributed to the desired effect of using the measurement setup to also stress the device. However, it cannot fully account for the observed behavior.

Figure 5.13: Charge pumping current $Icp$ for different pulse amplitudes as observed in constant base-level CP measurements with $VBase = 2V$ and a gradually increasing pulse amplitude $ΔVG = VBase − VG,low$ from $VG,low = − 1V$ down to $VG,low = − 17V$ . $Icp$ shows a significant hysteresis. If $Icp$ is evaluated at the falling pulse edge, the lower branch of the curve is traversed. Evaluation of the rising pulse edges gives the upper branch. However, the contribution of slow oxide states and an additional hysteresis (marked with $ΔIcp$ ) are clearly visible for increasing pulse amplitudes. This implies that depending on the pulse amplitude, $Icp$ will contain contributions of both, interface and oxide states. Provided only interface states are available, $Icp$ should be independent of the pulse amplitude (dashed line of $I it cp,0$ ).

5.5.1 Dependence on Gate Voltage Low-Level
5.5.2 Hysteresis due to Stress

[next] [prev] [prev-tail] [front] [up]