Investigation on the structural property of the sputtered hcp-phase boron nitride tunnel barrier for spintronic applications

Magnetic tunnel junctions (MTJs) have attracted significant attention in the past decades due to their promising application in next-generation ultra-high density and non-volatile memory devices.^1–4 The fundamental structure of a MTJ is two ferromagnetic layers separated by an insulating tunnel barrier (e.g. MgO).^5,6 When passing through the MTJ devices, the charge current will be polarized and a resulting spin current can be generated. Through the tunnel effect of the MTJ barrier, the spin-transfer-torque (STT) caused by the spin current can switch the magnetization of the free layer of the MTJ devices.⁷ To obtain large STT switching efficiency, maintaining a low resistance-area (RA) product is one of the most essential device characteristics. Currently, a MgO tunnel barrier is commonly used due to the matching of the lattice and band structure with the ferromagnetic electrodes.^5,6,8 However, when decreasing the thickness of the MgO tunnel barrier down to ∼1.0 nm to realize ultra-low RA products, the non-uniformity, defects, and pinholes of the MgO tunnel barrier are harmful and even fatal to device performances.⁹ In addition, the cubic-phase MgO tunnel barrier cannot match with the desirable hexagonal close-packed (hcp) phase ferromagnetic or antiferromagnetic electrodes for developing hcp-textured MTJs and antiferromagnetic devices. Recently, due to outstanding physical properties such as large bandgap (∼6 eV), high thermal and chemical stability, and negative electron affinity,^10–12 the 2D hcp-phase boron nitride (BN) material is predicted to be a very promising candidate of the tunnel barrier.^13–16 However, the BN tunnel barrier was developed primarily using exfoliation methods from a BN single crystal.^17,18 This process blocks the integration of the BN-based MTJ into current CMOS technologies. Thus, it is desirable to explore the industry-preferred sputtering process for fabrication of hcp-phase BN tunnel barriers for spintronic applications. In this work, the optimized experimental conditions are found to be the ratio of sputtering gas of Ar/N₂ ∼ 1:1 and substrate temperature at 340 °C. Then, we successfully grew a sputtered BN tunnel barrier with polycrystalline structure on the Co bottom electrode, which was characterized by the cross-sectional transmission electron microscopy (TEM). A high-temperature (850 °C) sputtering environment may improve the crystalline structure of the BN tunnel barrier following, as reported by P. Sutter et al.¹⁴ However, such high-temperature processing causes serious interdiffusion and makes the layers noncontinuous. The thermal treatment processing of the hcp-phase BN-based MTJ stacks and the resulting magneto-transport properties of the hcp-phase BN-based MTJ devices will be investigated in future work.

All the thin films were deposited using an eight-target ultra-high vacuum dc magnetron sputtering system with a base pressure better than 5×10⁻⁸ Torr. First, the BN single layer was deposited on a Copper (111) substrate to optimize the composition by tuning the ratio of Ar/N₂ sputtering gas. Then the MTJ stacks with structure of Ta (4)/Ru (20)/Co (7)/BN (2.1)/Co (7)/Ta (5) (thickness are in nanometers) were deposited on a Si/SiO₂ (300 nm) substrate to investigate the crystalline structure of the BN tunnel barrier. The Ta/Ru bilayer, serving as a buffer layer, was deposited at room temperature, then the substrate was heated to 340 °C to deposit the rest of the layers. The BN tunnel barrier layer was deposited with the Ar/N₂ gas at a pressure of 1.7×10⁻³ Torr with the ratio of Ar/N₂ ∼1:1, and all other layers were deposited with Ar gas at a pressure of 2×10⁻³ Torr. The composition of the BN layer was characterized by Auger electron spectroscopy (AES). The crystallographic structure of the MTJ stacks were measured by a Cu source X-ray diffractometer (XRD, Panalytical X’pert) and TEM (Tecnai T12 and F30).

In order to obtain the stoichiometric BN thin films, we first prepared a BN single layer on the Cu (111) substrates by changing the ratio of the mixed Ar/N₂ sputtering gas. Then the samples were measured by AES to identify the composition of the BN thin films. The results are plotted in Fig. 1. When only Ar sputtering gas is used, the AES peaks of the B and N elements measured are at 168 eV and 379 eV, respectively, as shown in Fig. 1(a), which is in agreement with those given in previous reports.^19,20 The composition of the BN thin film is calculated to be around 72:28 based on the AES spectra. Meanwhile, except for the B⁺ peak, we also observe the peak of the B element, as shown in the inset of Fig. 1(a). This evidence indicates that the B is rich in the BN thin film. Then, we tuned the ratio of mixed Ar/N₂ sputtering gas to 1:1 at a sputtering pressure of 1.7 $×$ 10⁻³ Torr. The two peaks around 168 eV and 379 eV of AES spectra are obtained as shown in Fig. 1(b). The composition of the BN thin film is then calculated to be approximately 50:50, and the peak of the B element disappears as shown in the inset of Fig. 1(b). From these results, we find that the mixed Ar/N₂ gas helps obtain the stoichiometric BN thin films.

To study the texture and crystalline structure of the BN tunnel barrier, we deposited a full MTJ stack with the structure of Ta (4)/Ru (20)/Co (7)/BN (2.1)/Co (7)/Ta (5) and a reference with the structure of Ta (4)/Ru (20)/Co (7) (thickness’ are in nanometers). Here, Co was selected as the ferromagnetic electrode because the Co layer (a=0.25017 nm) has a very small lattice mismatch to Ru (a=0.27059 nm) and hcp-BN (0.2504 nm) layers. The XRD measurements of the samples were carried out and the results are plotted in Fig. 2. As shown in Fig. 2, a (0002) peak at 42.15° of the Ru buffer layer and a hcp (0002) or fcc (111) peak at 44.25° of the Co layer are observed from the reference sample and the as-deposited MTJ stack, which means that the Ru buffer layer forms the hcp-phase. Meanwhile, the lattice constants of Ru and Co layers are calculated to be c ≈ 4.28 Å and ≈ 4.11 Å, respectively, which is similar to their bulk c-lattice constants. Meanwhile, the finite diffraction fringes around the (0002) peak of the Ru layer indicate that we obtained a highly textured Ru layer with smooth interfacial roughness. For the as-deposited MTJ stack, the peak at 44.25° of the Co layer is observed, which indicates the top Co layer is hcp-phase or fcc-phase on the BN tunnel barrier. For the MTJ stack annealed at 850 °C, the intensity of the (0002) peak of the Co layer is enhanced, indicating a crystalline improvement. However, since with increasing annealing temperature, both of the (0002) peak of Ru layer and the peak of Co layer shift to a larger 2θ angle (∼1.3°) and smaller 2θ angle, respectively, and the finite diffraction fringes disappear. The lattice constant along the c-axis is calculated to be ∼ 4.16 Å and 4.21 Å for the Co and Ru layers, respectively. This suggests the diffusion at the interface.^21,22 Because the BN layer is thin and likely possesses polycrystalline structure, we do not observe the identified peak of BN thin film in the XRD patterns.

To investigate the crystallography and interfaces of the MTJ sample, especially the BN tunnel barrier, cross-sectional TEM analysis was performed on the as-deposited MTJ sample. The results are shown in Fig. 3. As shown in Figs. 3(a) and 3(b), we found that the Ru buffer layer is textured and flat with a hcp-phase. While the Co layer deposited on the Ru buffer layer, it formed an island-like structure since the surface free energy of Co is higher than that of Ru¹⁴ and the substrate was heated during the deposition. Then the hcp-BN layer grown smoothly on the Co island surface as well. The reason is that the surface free energy of the BN layer is smaller than that of the Co layer. Figures 3(c) and 3(d) show the zoom-in TEM images of the Co/BN/Co structure. From these TEM images, we see that the bottom Co layer exhibits very strong texturing which is induced by the Ru buffer layer, and the BN layer shows a polycrystalline or an amorphous structure. The top Co layer followed the texture of the hcp-BN layer although it is not as high-textured as the bottom Co layer due to the polycrystalline structure of the BN layer.

In order to investigate the crystalline structure of the BN tunnel barrier with the high-temperature thermal treatment process, the MTJ stack was annealed at 850 °C for 2 mins using rapid thermal annealing (RTA) processing. Figure 4 shows the TEM images of the annealed MTJ stack. From Fig. 4(a), we can clearly see that the high-temperature annealing process damages the layered MTJ stack sample and makes the BN tunnel barrier noncontinuous. It is also found that the top Co layer shows serious diffusion. Co diffusion into the BN tunnel barrier in some regions is also observed, which is the underlying reason for why the high-temperature thermal treatment process ruins the BN tunnel barrier, as shown in Fig. 4(b). In addition, we can see that the BN tunnel barrier follows the texture of the upper Co layer after annealing in some regions (see the region labeled by the red circle) and some parts are amorphous (see the region labeled by the green circle), as shown in Fig. 4(c). From these results given above, we draw the conclusion that annealing at high temperature could improve the crystallinity at a certain degree, but may ruin the interface of the BN layer.

In summary, we have successfully developed a polycrystalline hcp-BN tunnel barrier by magnetron sputtering. And by optimizing the ratio of Ar/N₂ sputtering gas, we obtained stoichiometric BN thin films. The BN tunnel barrier possesses polycrystalline structure in the as-deposited MTJ stack. The high-temperature annealing process could improve the crystalline structure of the BN tunnel barrier; however, it causes serious interdiffusions among layers. Future works should focus on the optimization of the annealing temperature e as well as the study of magneto-transport properties. We believe that these results will lead to further study on the hcp-BN tunnel barrier for fabricating hcp textured perpendicular MTJs.

This work was carried out in the Characterization Facility, University of Minnesota, a member of the NSF-funded Materials Research Facilities Network (www.mrfn.org) via the MRSEC program and being partly supported by ASCENT, one of six of JUMP, a Semiconductor Research Corporation program, sponsored by MARCO and DARPA, grant reference number is 203278UMN Am 1.