Tunnel magnetoresistance sensors with CoFeBTa amorphous soft-magnetic sensing layer

Abstract

Tunnel magnetoresistance (TMR) magnetic sensors using CoFeB/MgO/CoFeB magnetic tunnel junction (MTJ) are promising for detection of magnetic field with various magnitude and frequency. The soft-magnetic property of the free layer (sensing layer) influences the sensing performance, which is particularly important for detection of small magnetic field such as biomagnetic field. NiFe (permalloy) is frequently used as a soft-magnetic sensing layer of TMR sensors. However, the large film roughness and the fcc 111-texture of the NiFe polycrystalline film tend to lower the TMR ratio. Amorphous ferromagnets such as CoFeSiB [1] have been developed as alternative materials for the sensing layer of TMR sensors. However, the soft-magnetic nature of CoFeSiB is prone to be degraded due to crystallization by annealing at > 300 °C [2], which is the typical annealing temperature for TMR sensors. In this work, we demonstrate an amorphous CoFeBTa material, whose crystallization temperature is ~500 °C [3], to the TMR sensing layer and investigated the magnetic, magnetoresistive and low-frequency noise properties.A top-pinned spin-valve structure of bottom lead / Ta(5) / CoFeBTa(20) / Ta(0.3) / CoFeB(3) / MgO(1.8) / CoFeB(2.5) / CoFeBTa(0.6) / CoFe(0.5) / Ru(0.8) / CoFe(3) / IrMn(8) / Ru(8) capping layer (nm in unit) was sputter-deposited [4]. The spin-valve film was annealed twice (two-step annealing) under magnetic fields in two directions that were orthogonal each other. The first annealing at 350 °C induces a uniaxial magnetic anisotropy in the CoFeBTa/Ta/CoFeB free layer in e.g. y-direction and the second annealing at 200 °C defines the pinning direction in the orthogonal direction (x-direction). The pattered TMR devices by photolithography were characterized by dc 4-probe method.Figure 1(a) shows the resistance (R)-magnetic field in x-direction (Hx) curve of a single TMR device (MTJ diameter: 50 μm). The TMR ratio was 160% and the resistance-area product was 40 kΩ μm2. A linear R-H transfer curve with a small hysteresis was realized in the intermediate magnetization state after the two-step annealing process. The maximum sensitivity (1/Rmin dR/dH) of ~70 %/mT was obtained at Hx = 1 mT. Figure 1(b) shows the noise voltage density (√SV) spectra of the TMR device at various magnetization states. For the noise measurements, bias voltage (Vb) was supplied using a 1.5 V dry cell battery and a ballast resistor. The magnitude of the noise at the intermediate magnetization state (Hx = 1 mT) was larger than those of the parallel and anti-parallel magnetization states, indicating that the 1/f noise is dominated by some distinct magnetic origins. The noise voltage density increased with increasing Vb, and the Hooge’s noise parameter αH was determined to be ~ 4×10-8 μm2 at Hx = 1 mT. The minimum detectivity (field-equivalent noise) defined by √SV/(1/Rmin dR/dH Vb) was 2.2 nT/√Hz at 10 Hz at Hx = 1 mT.For comparison, we fabricated a similar TMR sensor device with a NiFe(20 nm) soft-magnetic layer instead of CoFeBTa. The NiFe sensor showed larger detectivity of 4.5 nT/√Hz at 10 Hz than that of the CoFeBTa sensor mainly due to its larger 1/f noise. Thus, the CoFeBTa amorphous soft-magnetic layer has a clear advantage over the NiFe one for magnetic sensor applications.We also fabricated a full-bridge magnetic sensor using the CoFeBTa-based TMR devices. Each leg of the full-bridge is composed of 50 TMR devices connected in series as shown in Fig. 2(a). Figure 2(b) shows the output voltage (Vout) vs. H curve under various Vb. The non-linearity and hysteresis in the Vout-H curve can be attributed to variation of the R-H curve shape of the individual TMR devices of the array, which should be improved in the future. The bridge sensor showed an improved detectivity of 0.5 nT/√Hz at 10 Hz independent of Vb. **