Abstract

MgO-based magnetic tunnel junctions (MTJs) are a key component technology for spintronic devices such as hard drive heads and magnetic random access memory (MRAM). Moreover, there is a great deal of interest in utilizing MTJs in analog sensor applications such as magnetic field sensors [1], current sensors, and angle sensors because of the large tunneling magnetoresistance (TMR) effect of MTJs. In a new analog sensor application of MTJs, it has been reported that MTJs can also serve as a strain sensor by using a magnetostrictive sensing layer [2]. These MTJ strain sensors can be applied to various kinds of physical sensors by combination with micro-electro-mechanical-system (MEMS) technology, and we have recently reported a novel MEMS microphone utilizing MTJ strain sensors [3]. As seen from the above, the range of possibilities for analog sensor application using MTJs have greatly expanded, and there is strong demand for further improvements in the signal-to-noise ratio (SNR) of MTJ elements. To obtain high SNR in analog sensors, it is important to have not only a high TMR ratio, but also low 1/f noise. Annealing is effective for improving the TMR ratio because it enhances the crystallinity of the MgO barrier layer, and it has also been reported that annealing is effective for decreasing MgO barrier noise [4]. Thus, a higher annealing temperature is preferable in order to enhance these effects. However, in the case of MTJs with an exchange biased pinned layer, which is generally used in analog sensor applications, excessively high-temperature annealing causes degradation of the exchange bias field (H ex ). This degradation decreases the effective TMR ratio and increases the magnetic 1/f noise coming from pinned layer [5], thus creating an obstacle to the adoption of high-temperature annealing. In this study, we aimed to improve the annealing stability of the exchange biased pinned layer in MgO-MTJ for analog sensor applications. We therefore focused on enhancing the crystallinity of the anti-ferromagnetic IrMn layer and investigated the effect of an underlayer structure on the crystallinity of the IrMn layer and annealing stability of the exchange biased pinned layer. We fabricated MTJs consisting of underlayer/IrMn7/CoFe2.5/Ru0.9/CoFeB3/MgO1.8/CoFeB4/Cu1/Ta2/ Ru20 (layer thicknesses shown in nanometers). We prepared two kinds of underlayer structures: a generic Ta1/Ru2 underlayer (TR-UL); and an improved Ta1/Ru2/Ta2/Ru2 underlayer (TRTR-UL). Fig. 1 (a) and (b)show the R-H curves for the MTJ with TR-UL (TR-MTJ) and MTJ with TRTR-UL (TRTR-MTJ) annealed at $380\,^{\circ}\mathrm {C}$. As can be seen in Fig. 1, the TRTR-MTJ exhibits a larger H ex and larger TMR ratio compared with those of TR-MTJ, We also evaluated the noise level of these MTJs, with the Hooge parameter α estimated by fitting the frequency dependence of the noise level measured at 200 Oe. The estimated α values for both MTJs are shown in Fig. 1(a)and (b). The value of α at 200 Oe for TRTR-MTJ is lower than that for TR-MTJ. Therefore, the magnetic 1/f noise is clearly improved by using TRTR-UL thanks to the high annealing stability of the exchange bias field. To clarify the microstructure of both MTJs, cross-sectional high-resolution transmission electron microscopy (HRTEM) and electron energy-loss spectroscopy (EELS) were performed. Insets of Fig. 1(a)and (b)show the HRTEM and EELS results for TR-MTJ and TRTR-MTJ annealed at $420\,^{\circ}\mathrm {C}$, solid lines in the figure show the EELS intensity of Mn. These results confirmed that the Mn diffusion is quite different between TR-MTJ and TRTR-MTJ. For TR-MTJ, Mn diffused from the IrMn layer to the pinned layer, whereas for TRTR-MTJ, there is almost no Mn diffusion from the IrMn layer to the neighboring layer. To investigate why TR-MTJ and TRTR-MTJ exhibited different Mn diffusion, the crystallinity of IrMn for both MTJs was analyzed by X-ray diffraction measurement. As a result, the crystallinity of IrMn on TRTR-UL was found to be clearly improved compared with that on TR-UL. Furthermore, we investigated the detailed structure of TR-UL and TRTR-UL. TEM measurement and fast Fourier transform measurement (FFTM) analysis were used to confirm the detailed d spacing of each layer in TRTR-UL (Fig. 2). From Fig. 2, it was successfully confirmed that the first Ta layer and second Ta layer in TRTR-UL have different d-spacing. The second Ta layer deposited on Ru layer has the d-spacing corresponding to the (110) textured α-Ta structure and highly crystallinity, whereas the first Ta layer does not have the component of (110) textured α-Ta structure. These results indicate that Ta-on-Ru structure is responsible for the high crystallinity of IrMn layer disposed on TRTR-UL, resulting in the high annealing stability thanks to small Mn diffusion. Therefore, Ta/Ru/Ta/Ru structure is promising as an underlayer for realizing a high SNR in MTJ analog sensor applications.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call