“…1 shows the bias-stress results, performed in devices with different channel lengths L in the SH regime under constant power density (0.04 mW/µm 2 ) for the same V g − V t = −12.2 V and high V ds and for a bias-stress time of up to 1000 s. As can be noted, the V T shift after 1000 s of bias stress was increasing for increasing L (∆V T = −4.5, −5.9, and −8.7 V for L = 1, 2, and 6 µm, respectively), and the devices with L = 2 and 6 µm show a g m overshoot, after bias stress, which instead is not observed in the short-channel device (L = 1 µm). It should be pointed out that, when performing a thermal annealing at T > 180 • C before bias-stress experiments, the g m overshoot is not observed [3]. In addition, we have also reported that bias stress under NBTI conditions can also produce a g m overshoot, similar to the g m overshoot induced by bias stressing in the SH regime [4].…”