Magnetic Viscosity in High Precision Magnetoresistive Field Sensors

Abstract

Magnetic viscosity, the long time-scale magnetization drift in response to changes in applied field, can directly impact the accuracy of high sensitivity tunneling magnetoresistance (TMR) based field sensors operating with large dynamic field ranges or utilizing long time scales. Recent studies have demonstrated benefits arising from an understanding of magnetic viscosity in the context of field sensing; for instance, as a source of spurious signals from buried ferrite during landmine detection by induction sensors [1], as a tool providing insight into magnetization dynamics [2-5], and even as a novel immunoassay characterization technique for biochemical detection through so-called magnetorelaxometry [6,7]. Although the mechanism of magnetic viscosity has been well described [8,9], quantifying the impacts of the effect in sensor applications and establishing relationships between those impacts and tunable material properties has remained under explored. Therefore, in this work we characterize the magnetic viscosity in a variety of commercial and custom magnetic tunnel junction (MTJ) based sensors, as well as explore the potential impacts of integration with flux concentrator shielding. Utilizing accessible electrical measurements of the TMR or sensor output, a wide range of viscosity behavior is revealed. The results in commercial sensors demonstrate order-of-magnitude differences ranging from 0.2 – 2 μT in the effective sensor output field viscosities, with accuracy implications due to the hysteric nature of the viscous drift in the sensor output. To elucidate underlying differences in the selected commercial sensors, we fabricate and compare custom MTJs with variable free layer thicknesses tFL as well as patterned devices utilizing a variety of flux concentration shielding (with up to 30x flux concentration factors). Fig. 1(a) presents an illustrative case of the magnetoresistance vs applied field for one of the custom MTJs where tFL = 4 nm for a biased sensing layer. For each measurement, the device is first initialized in a saturating field μ0H0 of approximately 30 mT. The field is then abruptly reduced to the measurement field μ0Ha, ranging from -25 to 25 mT. The time-dependent rate of change of the MTJ’s resistance for such a measurement with the specified value of Ha is shown in Fig. 1(b), illustrating that the observed drift of the device resistance is logarithmic in time and strongly field dependent as expected. These measurements are then used to extract the magnetoresistance’s viscosity parameter. The patterned MTJs studied show a 3x increase in sensitivity between tFL = 4 to 20 nm but a corresponding 43% decrease in the magnetic viscosity parameter [Fig. 2(a-b)]. Relationships between the viscosity, magnetic active volumes, and anisotropy will be discussed. These results highlight simple design pathways useful to consider when attempting to enhance or mitigate viscous output effects in high sensitivity MTJs. In addition, the viscosity parameter in certain fabricated MTJs is found to exhibit a strong asymmetric field dependence that is not observed in the sensitivity [Fig. 2(a-b)], highlighting complexities of the reversal mechanism, as well as the deeper insight into magnetization dynamics gained from viscosity characterization. **