C low - Frequency Ultrasensitive Magnetic detectors CNF Project

Figure 1: Effect of annealing on tunneling magnetoresistance ratio. Low-noise magnetic sensors (noise spectral density ~ pT/Hz1⁄2) based on magnetic tunnel junctions can be integrated to provide unprecedented clinical information, viz., room-temperature magnetocardiography [1]. The sensors comprise a tunnel junction, a “flux concentrator,” and a mechanism for chopping the input signal with a microelectromechanical actuator to reduce 1/f noise. An array of sensors assembled on a planar substrate along with suitable digital signal processing modules could yield a system that is capable of imaging, both in time and space, the magnetic field generated by depolarization/repolarization currents in the chest cavity during a cardiac cycle. Summary of Research: Magnetic tunnel junctions are routinely used as ultrasensitive magnetic sensors, for example in computer hard disk drives. However their low-frequency performance is limited by low-frequency noise, i.e., 1/f noise. We investigated 1/f noise in high-performance MgO magnetic tunnel junctions (MTJs) with a tunneling magnetoresistance (TMR) of 160%, and examined the influence of annealing and MTJ size on the noise. The results show that the annealing process not only dramatically improves the TMR, but can also strongly decrease the MTJ noise. Junctions were prepared in a deposition system dedicated to synthesis of magnetic nanostructures. A typical MTJ stack had the structure Si/Ta(5)/Ru(15)/Ta(3)/IrMn(10)/Co60Fe20B20(3)/ MgO(2.2)/Co60Fe20B20(3)/Ta(8)/Ru(7), where the thicknesses in parentheses are in nm. The films were annealed at various temperatures in order to improve the quality of the interface at the tunneling barrier and thereby maximize the tunneling magnetoresistance. Figure 1 shows the relation between TMR and anneal temperature, showing that a value as high as is achieved for Tanneal = 380°C [2]. The low-frequency noise of these tunnel junctions is also reduced at the optimum annealing temperature. We observe a current-independent noise voltage that is not very frequency dependent, and find that the 1/f noise is approximately linearly dependent on current indicating an origin in tunneling fluctuations. Figure 2 shows the frequency and current dependence of the noise voltage for an unannealed and optimally-annealed device. The noise level is usefully characterized by the “Hooge parameter” defined as where A is the area of the tunnel junction. At 10 Hz, the Hooge parameter in the state with the magnetizations parallel (low resistance) decreases from 3.8 × 10-8 μm2 for the unannealed device to 7.1 × 10-9 m2 for the optimally annealed device—a fivefold decrease. We attribute this decrease to a reduction in the number of imperfections within the barrier layer, reducing low frequency occupation variations that alter the tunneling probability roughly fourfold. the noise is found to depend on the applied magnetic field, indicating that magnetic noise also contributes somewhat to the overall noise. The Hooge parameter indicates that the noise can be reduced by increasing the area Figure 2: Power spectral density of the noise observed in (left) an unannealed device and (right) a device annealed to achieve maximum TMR. The raw voltage noise is lower, and since the sensitivity is increased greatly (Figure 1) the effective magnetic noise is much lower.