Pulsed spin-torque switching has been studied using single-shot time-resolved electrical measurements in perpendicularly magnetized magnetic tunnel junctions as a function of pulse amplitude and junction size in 50 to 100 nm diameter circular junctions. The mean switching time depends inversely on pulse amplitude for all junctions studied. However, the switching dynamics is found to be strongly dependent on junction size and pulse amplitude. In 50 nm diameter junctions the switching onset is stochastic but the switching once started, is fast; after being initiated it takes less than 2 ns to switch. In larger diameter junctions the time needed for complete switching is strongly dependent on the pulse amplitude, reaching times less than 2 ns at large pulse amplitudes.Anomalies in the switching rate versus pulse amplitude are shown to be associated with the long lived (> 2 ns) intermediate junction resistance states. PACS numbers:1 arXiv:1610.09710v1 [cond-mat.mes-hall]

No abstract

No abstract

No abstract

Spin-transfer magnetic random access memory is of significant interest for cryogenic applications where a persistent, fast, low-energy consumption and high device density is needed. Here we report the low-temperature nanosecond duration spin-transfer switching characteristics of perpendicular magnetic tunnel junction (pMTJ) nanopillar devices (40 to 60 nm in diameter) and contrast them to their room temperature properties. Interestingly, at fixed pulse voltage overdrive the characteristic switching time decreases with temperature, in contrast to macrospin model predictions, with the largest reduction in switching time occurring between room temperature and 150 K. The switching energy increases with decreasing temperature, but still compares very favorably to other types of spin-transfer devices at 4 K, with < 300 fJ required per switch. Write error rate (WER) measurements show highly reliable (WER ≤ 5×10 -5 with 4 ns pulses at 4 K) demonstrating the promise of pMTJ devices for cryogenic applications and routes to further device optimization.