When I measured home-made, short channel (L10 micron) bulk CMOS, anomalous behavior as shown in blue plots in the attached file was observed.
Personally, artifacts of measurements and inhomogeneity of doping processes are possible.
Have anyone experienced such behavior?
FYI, I hope you will find following information useful (I am afraid full data were unavailable because of contracts).
Each device has four pads for drain, source, bulk, gate electrodes.
Measurements were done using 4200SCS mounted on a manual prober having four probes.
Halo doping was employed for the devices.
Long channel devices (L>0.6 micron) as well as narrow channel devices (W= 0.5 micron) were negative (did not show the behavior /the behavior was not observed).
Positive (the behavior was observed) for Vd = 1 V and negative for Vd = 50 mV.
Use of another 4200SCS did not change results.
Switching SMUs for biasing was... hard to describe... it was positive for some cases and negative for the other even using the same devices. For example, blue and red lines in the attached file were obtained from the same device with using different measurement configurations.
Switching probes did not change results.
Changing current ranges for measurements did not change results.
Changing hold, and delay times (from 0 ms up to 20 ms) for measurements did not change results.
Changing Vg sweep direction did not change results.