1. Introduction

Currently there is intense interest in MTJ with PMA for its potential applications in magnetic random access memory and spin-orbit torques devices^[1-10]. Among the studies on perpendicular magnetic tunnel junctions (p-MTJ) to date, rare earth transition metal alloys TbFeCo and GdFeCo have been used as the ferromagnetic electrodes^{[7, 11, 12]}. In addition to poor oxidation resistance, these alloys suffer from the drawback of losing perpendicular anisotropy when annealed above 200 ℃^[13-15]. The recent discovery of ultrathin CoFeB/MgO/CoFeB tunnel junctions with perpendicular magnetic anisotropy, satisfying the criteria of high tunneling magnetoresistance ration, i.e. high anisotropy constant and low switching current, has aroused a great deal of interest as a promising candidate for spintronics materials^{[16, 17]}. As pointed out by Ikeda $\textit{et al}$.^[18] for the PMA of CoFeB films, the effects of the post annealing temperature, CoFeB composition, and capping layer on the film performance were carefully investigated^[19-23], but there are few reports on the origin of PMA in Ta/CoFeB/MgO structure.

It is generally admitted that the PMA is governed by interfacial bounding states between Co(Fe) and oxygen^{[18, 24]}, which are known to experience a more drastic change in annealing^{[25, 26]}. Worledge $\textit{et al}$.^[27] found that the PMA energy from Ta/CoFeB/MgO is larger than Ru/CoFeB/MgO and concluded that PMA exists both in Ta/CoFeB and CoFeB/MgO interfaces. In this article, we have investigated the origin of PMA in Ta/CoFeB/MgO structure with the thickness of the CoFeB layer fixed to 1.2 nm, which is a typical thickness for CoFeB films exhibiting PMA demonstrated by several groups^{[19, 21, 23]}. Our experimental results show that the PMA is originated from the interface of CoFeB/MgO, and enhanced by Ta film obviously.

2. Experiments

All samples were prepared on an amorphous SiO$_{2}$ by magnetron sputter. To obtain a smooth surface, the base pressure was better than 6.5 $\times$ 10$^{-4}$ Pa. Three different structures have been studied, including sample S1: Substrate/Ta (5 nm)/Ru(10 nm)/Ta(5 nm)/CoFeB(1.2 nm)/MgO(2 nm)/Ta (5 nm), sample S2: Substrate/Ta(5 nm)/Ru(10 nm)/Ta (5 nm)/CoFeB(1.2 nm)/Ta (5 nm), and sample S3: Substrate/Ta (5 nm)/Ru(10 nm)/CoFeB(1.2 nm)/MgO(2 nm)/Ta(5 nm).

The CoFeB electrode layer was deposited by direct current (DC) sputtering with Co$_{20}$Fe$_{50}$B$_{30}$ (at.%) alloy target, and the MgO layer was deposited by radio-frequency sputtering with MgO target. The depositions of Ta, Ru, CoFeB and MgO were performed with the powers of 100, 100, 100, and 50 W, respectively. The experiments were implemented in Ar ambience with the flow rate of 20 sccm as working gas to bombard the target, the working argon pressure was 0.267 Pa, the deposition processes were operated under room temperature, and then all samples were annealed in a vacuum furnace with a base pressure of 8.0 $\times$ 10$^{\mathrm{-4}}$ Pa and a perpendicular magnetic field of 1 T for 1 h at a temperature of 350 ℃. Magnetic properties were studied using a Vibrating sample magnetometer (VSM), and the elementary compositions were analyzed by X-ray photoelectron spectroscopy (XPS).

3. Results and discussion

As is well known, the effective magnetic anisotropy energy $K_{\rm eff}$ is expressed by Eq. (1)^[28].

$\begin{equation} K_{\rm eff} =(K_{\rm V} -2\pi M_{\rm S}^{2})+2K_{\rm S} /t, \end{equation}$

(1)

where $K_{\rm V}$ is the volume magnetic anisotropy, $K_{\rm S}$ is the interfacial magnetic anisotropy, $t$ is the thickness of CoFeB and $M_{\rm S}$ is the saturation magnetization. The value of $K_{\rm eff}$ is taken from the contested result between $K_{\rm {V}}$ and $K_{\rm S}$. From Ref. [29], $K_{\rm eff}$ is equal to $H_{\rm k}M_{\rm s}$/2 in numerals, and $H_{\rm k}$ is the anisotropy field of the hard axis, which is negative if the film has the In-plane magnetic anisotropy (IMA) and positive if the film has PMA. $H_{\rm k}$ can be written by Eq. (2)^[30]:

$\begin{equation} \label{eq1} H_{\rm K} =2\int_0^{H_{\rm {u}p} } {[m_{\rm {e}a} (H)-} m_{\rm ha} (H)]{\rm d}H, \end{equation}$

(2)

where $m=M/M_{\rm S}$, which is the specific magnetization, $m_{\rm {ea}}$ is the specific magnetization intensity along the easy axis, and $m_{\rm ha}$ is the specific magnetization intensity along the hard axis, the upper integration boundary $H_{\rm up}$ was chosen to be higher than the saturation field of the hard axis. As $| K_{\rm eff}|=H_{\rm k}M_{\rm s}$/2, according to Eq. (2),

$\begin{equation} \label{eq2} \left| {K_{\rm eff} } \right|=\int_0^{H_{\rm {u}p} } {[m_{\rm {e}a} (H)-} m_{\rm ha} (H)]{\rm d} H\cdot M_{\rm S} . \end{equation}$

(3)

Therefore, $K_{\rm eff}$ can be determined as the magnetization area difference between out-of-plane and in-plane. When the easy axis is along the in-plane, the $K_{\rm eff}$ is negative and corresponds to IMA while the positive $K_{\rm eff}$ corresponds to PMA when the easy axis is along the out-of-plane.

Exemplary results from the sample S1 are shown in Fig. 1, where the red line and blue line draw up the out-of-plane and in-plane magnetic hysteresis loops respectively. In Fig. 1, the in-plane magnetic hysteresis loop is hard to reach saturation, and $K_{\rm eff}$ is equal to $1.47 \times 10^{\mathrm{6}}$ erg/cm$^{\mathrm{3 }}$ for the sample S1. Magnetic hysteresis loop and $K_{\rm eff}$ indicate that S1 exhibits PMA.

Fig. 2 shows the in-plane and out-of-plane magnetic hysteresis loops for S2, which are hard to reach saturation, while $K_{\rm eff}$ is negative and equal to $-1.287 \times 10^{6}$ erg/cm$^{3}$. The PMA could be stronger than that in the sample of S1 due to Ta/CoFeB and CoFeB/Ta existing in S2 if the PMA originated from the Ta/CoFeB interface. But the truth is just the opposite and the PMA has disappeared. It is suggested that the PMA does not originate from the interface of Ta/CoFeB, and the MgO layer plays an important role on PMA. In the deposition process, Co and Fe atoms are over-oxidized by O atoms, which are excited from the MgO target. After the annealing process, all the Co-oxide and most of Fe-oxide are reduced to their metallic states, and the PMA is realized due to the appropriate hybridization of the Fe-3d and O-2p orbits at the ferromagnet/oxide interface^[17]. Fig. 3 shows the XPS spectra of Fe peaks at the interface of CoFeB/MgO for S1. The 2p$^3$ and oxides states of Fe atoms exist in 706.7 eV and 710-711.5 eV respectively. FeO and Fe$_{2}$O$_{3}$ coexist in 710-711.5 eV, the percentage of which is 50.9% by calculation. In the annealing process, MgO film is crystallized gradually, inducing the crystallization of CoFeB film and the atoms arranged regularly at the interface of CoFeB/MgO, as shown in Fig. 4(a). Both Fe-oxide and atom rank are critical reasons for PMA. However, in the sample of S2, the PMA has disappeared because the interfaces of Ta/CoFeB/Ta are in amorphous and metallic contact, as shown in Fig. 4(b).

Fig. 5 shows the $M-H$ curves for the sample of S3 with the magnetic field in-plane and perpendicular direction, respectively, and $K_{\rm eff}$ is equal to 0.797 $\times$ 10$^{6}$ erg/cm$^{3}$. Compared with S1, there is a perpendicular magnetization easy axis existing in the sample of S3, and the effective perpendicular magnetic anisotropy field decreases as CoFeB film is deposited on Ru directly. It indicates that the interface of Ta/CoFeB is beneficial to enhance the effective perpendicular anisotropy, even though it is not the origin of PMA, because Ta film has better absorption ability with B atoms from CoFeB film^[31], and then promotes the crystallization of CoFeB. Combining with MgO film and the annealing process, the PMA of sample S1 is superior to the other two samples.

4. Conclusion

In summary, three different structures have been presented to study the origin of PMA in the CoFeB/MgO/CoFeB system. During the deposition and annealing process, Fe-oxide is formed at the interface of CoFeB/MgO while Ta film absorbs B atoms from CoFeB film, so both of them are enhancing the PMA of the Ta/CoFeB/MgO structure. The comparisons of magnetic properties in Ta/CoFeB/MgO, Ta/CoFeB/Ta and Ru/CoFeB/MgO structures demonstrate that the PMA is originated from the interface of CoFeB/MgO rather than Ta/CoFeB, while the Ta film is helpful to enhance the PMA of the CoFeB/MgO/CoFeB structure.