In this letter, we present a study of optimized TMR magnetic field sensors as a function of voltage bias. The 1/f low-frequency noise is quantified by the Hooge-like parameter {alpha} which allows to compare the low-frequency behavior of various TMR sensors. The sensitivity as well as the detectivity of the sensor are characterized in the parallel state and at 0 mT. We observe that the sensitivity shows a strong voltage dependence and the noise presents an unexpected decrease, not anticipated by the Hooges law. Moreover, surprisingly, an almost stable detectivity (140-200 nT/sqrt(Hz) at 10 Hz and 15-20 nT/sqrt(Hz) at 1 kHz) as a function of the bias voltage is observed, tending to highlight that the variation of sensitivity and noise are correlated. Even if the I-V curves are strongly non-linear and reflect the different symmetries of the conduction bands channels, the variations in sensitivity and noise seems to depend mainly on the distortion of the MgO barrier due to bias voltage. With a simple model where the normal noise and sensitivity of the TMR sensors are modified by an element having no noise and a parabolic conductance with voltage, we describe the behavior of noise and sensitivity from mV to V.
Download