Magnetization dynamics in racetrack memory

Abstract

Various devices have been proposed which use magnetic domain walls (DWs) in nanosized magnetic structures to perform logic operations or store information. In particular in ‘Racetrack memory’ bits of information represented by DWs are shifted in a magnetic wire to be stored. For these memory and logic devices to be successful, great control of DW motion is of vital importance. In cooperation with IBM’s Almaden research laboratory a pump-probe Kerr magnetooptical scanning microscope has been developed. In order to control DW injection, motion and reset, magnetic fields have to be applied locally on the nanowire. For this a special Damascene CMOS chip has been fabricated at the 200 mm wafer facility at IBM Microelectronics Research Laboratory (MRL). Probing of the local magnetization is done with a focused pulsed laser spot of 400 nm diameter where the polarization rotation caused by the Kerr effect is measured after reflection. In order to achieve optimal focusing a perpendicular incident laser beam is focused with a high numerical aperture objective. Synchronized ‘pumping’ in this scheme is achieved by successively: 1 injecting a DW; 2 propagate the DW down the nanowire with either current through or an applied field pulse over the nanowire; 3 and finally resetting the whole nanowire to its original magnetization by applying a large field together with the injection of an opposite magnetic domain. With this setup field and current induced DW motion is studied in permalloy nanowires ranging in width from 200 to 700 nm and thickness of 20 nm. For control of DWs in Racetrack memory it is important to understand the different mechanism for driving a DW already in motion (dynamic) and driving a DW that is currently at rest (static). The propagation field, the minimum field below which no DW motion takes place, is measured for both dynamic DWs and static DWs. It is found that Static DWs require a much higher field than DWs already in motion. A model is build where this effect is related to the wire roughness, successfully describing the existence of a propagation field, the difference between both propagation fields and a specific effect related to the method of injection. For Racetrack memory to be successful the critical current needs to be small (the current needed to move a DW solely by current) and the DW velocity high. Much of the influence of intrinsic magnetic properties of materials on DW dynamics is unknown. One important property affecting DW velocity and possibly also the critical current is Gilbert damping. Gilbert damping in permalloy can be tuned by doping the nanowires with osmium. This is used to prepare a sample series with increasing Gilbert damping. Measurement of the field induced DW velocity revealed a profile well known that includes the Walker breakdown (a maximum field where further increasing field strength does not further increase the DW velocity). From this profile the dependence of the Walker breakdown, DW mobility and maximum DW velocity on Gilbert damping has been determined. With the same sample series also current induced field assisted DW motion has been measured. Current induced DW motion is known to be driven by two effects: adiabatic and ballistic- spin momentum transfer (SMT) which relative contribution is parameterized by beta in the Landau Lifshitz Gilbert equation (LLG). Measurement of DW velocity depending on current density revealed the relative contribution of the two SMT schemes. Also the influence of Gilbert damping on the relative contribution of both schemes has been explored. A pronounced dependence of the measured spin torque efficiency on osmium concentration was found. This result may be interpreted as a sign that the intensively debated ratio ? / ? is far from constant over the range of ? studied.