Validity of the thermal activation model for spin-transfer torque switching in magnetic tunnel junctions

We have performed spin-transfer torque switching experiments with a large number of trials (up to 10(7) switching events) on nanoscale MgO magnetic tunnel junctions in order to test the validity and the limits of the thermal activation model for spin-torque-assisted switching. Three different methods derived from the model ("read disturb rate," " switching voltage versus pulse duration," and " switching voltage distribution" measurements) are used to determine the thermal stability factor and the intrinsic switching voltage. The results obtained from the first two methods agree well with each other as well as with values obtained from quasistatic measurements, if we use only the data for which the voltage is smaller than approximately 0.8 of the intrinsic switching voltage. This agreement also shows that, in our samples, in the low voltage region, the influence from other factors contributing to the switching (such as current-induced heating and field-like torque) is negligible. The third method (switching voltage distribution measurements) yields incorrect values for the time-scales (< 1 mu s) at which the experiments are performed. Macrospin simulations confirm our findings that the model must be applied only in the low voltage limit, and that in certain devices this limit can extend up to about 0.9 of the intrinsic switching voltage. [doi:10.1063/1.3562136]