We model "soft" error rates for writing (WSER) and for reading (RSER) for perpendicular spintorque memory devices by solving the Fokker-Planck equation for the probability distribution of the angle that the free layer magnetization makes with the normal to the plane of the film. We obtain: (1) an exact, closed form, analytical expression for the zero-temperature switching time as a function of initial angle; (2) an approximate analytical expression for the exponential decay of the WSER as a function of the time the current is applied; (3) comparison of the approximate analytical expression for the WSER to numerical solutions of the Fokker-Planck equation; (4) an approximate analytical expression for the linear increase in RSER with current applied for reading; (5) comparison of the approximate analytical formula for the RSER to the numerical solution of the Fokker-Planck equation; and (6) confirmation of the accuracy of the Fokker-Planck solutions by comparison with results of direct simulation using the single-macrospin Landau-Lifshitz-Gilbert (LLG) equations with a random fluctuating field in the short-time regime for which the latter is practical.

We have performed spin-transfer torque switching experiments with a large number of trials (up to 107 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 (<1μs) 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.

We have measured the relaxation time of a thermally unstable ferromagnetic nanoparticle incorporated into a magnetic tunnel junction (MTJ) as a function of applied magnetic field, voltage V (-0.38 V < V < +0.26 V), and temperatures (283 K< T< 363 K) . By analyzing the results within the framework of a modified Néel-Brown formalism we determine the effective attempt time of the nanoparticle and also the bias dependences of the in-plane and out-of-plane spin transfer torques. There is a significant linear modification of the effective temperature with voltage due to the in-plane torque and a significant contribution of a "field like" torque that is quadratic with voltage. The methods presented here do not require complicated models for device heating or calibration procedures, but instead directly measure how temperature, field, and voltage influence the energy landscape and thermal fluctuations of a two-state system. These results should have significant implications for designs of future nanometer-scale magnetic random access memory elements and provide a straightforward methodology to determine these parameters in other MTJ device structures.

We demonstrate a simple low-power, magnetic sensor system suitable for high-sensitivity magnetic-field mapping, based on solid-state magnetic tunnel junction devices with minimum detectable fields in a 100 pT range at room temperature. In this paper, we discuss a method that uses multilayer thin films to improve the performance of the soft ferromagnetic layer in magnetoresistive sensor applications, by reducing the coercivity and/or improving the reversibility. We have used it in the design of our new magnetic sensor. This sensor has a sensitivity as high as 750%/mT. The magnetic sensor only dissipates 1 mW of power while operating under an applied voltage of 1 V.

No abstract