Double MgO-based Perpendicular Magnetic-Tunnel-Junction Spin-valve Structure with a Top Co2Fe6B2 Free Layer using a Single SyAF [Co/Pt]n Layer

A new perpendicular spin-transfer-torque magnetic-tunnel-junction (p-MTJ) spin-valve was developed to achieve a high tunneling magnetoresistance (TMR) ratio. It had a double MgO-based spin-valve structure with a top Co2Fe6B2 free layer and incorporated a single SyAF [Co(0.4 nm)/Pt(0.3 nm)]3 layer and a new buffer layer of Co(0.6)/Pt(0.3)/Co(0.4). It had a TMR ratio of 180% and anisotropy exchange field (Hex) of 3.44 kOe after ex-situ annealing of 350 °C for 30 min under a vacuum below 10−6 torr and a perpendicular magnetic field of 3 tesla, thereby ensuring a memory margin and avoiding read disturbance failures. Its high level of performance was due to the face-center-cubic crystallinity of the MgO tunneling barrier being significantly improved by decreasing its surface roughness (i.e., peak-to-valley length of 1.4 nm).

However, a double MgO-based p-MTJ spin-valve structure with a top CoFeB free layer remains a challenging target, because the roughness of the MgO tunneling barrier lower the TMR ratio [20][21][22][23][24] . Thus, in the study reported here, we designed a new double MgO-based p-MTJ spin-valve structure with a top CoFeB free layer using a single synthetic anti-ferromagnetic (SyAF) [Co/Pt] n layer instead of a double SyAF [Co/Pt] n layer that greatly reduces the roughness of the tunneling barrier, as shown in Fig. 1. First, we investigated the dependency of the TMR ratio on the body-centered-cubic (b.c.c) W bridge layer thickness for a double MgO-based p-MTJ spin-valve with a top Co 2 Fe 6 B 2 free layer using a single SyAF [Co/Pt] n layer [ Fig. 1(b)] and compared it with that of a double MgO-based p-MTJ spin-valve with a top Co 2 Fe 6 B 2 free layer using a double SyAF [Co/Pt] n layer [i.e., the conventional SyAF [Co/Pt] n layer: Fig. 1(a)]. Second, to understand the magnetic properties that differentiate the TMR ratio, we analyzed the static spin-torque-transfer behavior of the two different p-MTJ spin-valves by using vibrating sampling magnetometer (VSM). Third, to determine the reason for the TMR ratio difference, we observed the face-centered-cubic (f.c.c) crystallinity of the MgO tunneling barrier and Co 2 Fe 6 B 2 pinned layer of both spin-valves by using cross-sectional high-resolution-transmission-electron-microscopy (x-HRTEM). Finally, to investigate the dependencies of the current-vs.-voltage (I-V) curve, parallel-to-antiparallel switching voltage (V P-to-AP ), antiparallel-to-parallel switching voltage (V AP-to-P ), reading current in the low-or high-resistance state (I P or I AP ), and the normalized TMR ratio-vs.-voltage curve on the structure of SyAF [Co/Pt] n layer (i.e., a single or double), we fabricated p-MTJ spin-valve cells with a 34-nm diameter bottom electrode and 60μm-diameter top-electrode (see Supplement 1).

Results
TMR ratio of p-MTJ spin-valves. The dependency of the TMR ratio on the W bridge layer thickness (t W ) was investigated as a function of the structure of SyAF [Co/Pt] n layer, as shown in Fig. 1(c). For the double MgObased p-MTJ spin-valve with a top Co 2 Fe 6 B 2 free layer using the double SyAF [Co/Pt] n layer (i.e., the conventional SyAF [Co/Pt] n layer), there was a slight increase in the TMR ratio for t W values up to around 0.3 nm, since the ferro-coupling strength slightly increased with t W . The TMR ratio abruptly decreased when t W exceeded 0.36 nm, because the ferro-coupling strength between the upper [Co/Pt] 3 SyAF layer and the Co 2 Fe 6 B 2 pinned layer weakened abruptly. Thus, the maximum TMR ratio was only about 156% at a t W of 0.3 nm. On the other hand, the TMR ratio of the spin-valve with a top Co 2 Fe 6 B 2 free layer using a single SyAF [Co/Pt] n layer slightly increased up to a t W of 0.24 nm and then slightly decreased between t W values of 0.24 nm and 0.36 nm. The TMR ratio then rapidly decreased for t W values greater than 0.36 nm. Thus, the TMR ratio reached 180% at a t W of 0.24 nm. The dependency of the TMR ratio on t W for the p-MTJ spin-valve using the single SyAF [Co/Pt] n layer was similar to that of the spin-valve using the double SyAF [Co/Pt] n layer. However, the TMR ratio of the spin-valve using the single SyAF [Co/Pt] n layer (i.e., 180%) was much higher than that of the one using the double SyAF [Co/Pt] n layer (i.e., 156%). In particular, this means that the Co (0.6 nm)/ Pt (0.3 nm)/Co (0.4 nm) bridging buffer layer shown in Fig. 1   In the case of the spin-valves using the double SyAF [Co/ Pt] n layer, at t W of 0.16 nm, the spin-electron direction of the upper SyAF [Co/Pt] 6 layer ferro-coupled with the Co 2 Fe 6 B 2 pinned layer rotated from upward to downward when the external perpendicular magnetic-field (H) was scanned from +4 to +1.5 kOe, as shown in Fig. 2(a). Moreover, the spin-electron direction of the Co 2 Fe 6 B 2 free layer rotated from upward to downward when H was scanned from + 0.2 to −0.2 kOe, showing that free layer had good perpendicular magnetic characteristics (i.e., the M-H loop showed good squareness) and its magnetic moment was ~0.2 memu, as shown in the inset of Fig. 2(a). The spin-electron direction of the lower SyAF [Co/ Pt] 3 layer then rotated from upward to downward and the magnetic moment of the spin-valve saturated at 0.91 memu when H was scanned from −1.5 to −4 kOe. As a result, at a t W of 0.16 nm, the spin-valve had an anisotropy exchange field (H ex ) of 2.36 kOe ( Fig. 2(a)). Next, when t W was changed from 0. 16   lower SyAF [Co/Pt] 6 layer considerably improved (compare the red boxes in Fig. 2(a) and (b)). In addition, the squareness of the M-H loop of the Co 2 Fe 6 B 2 free layer slightly improved, resulting in an increase of the TMR ratio from 143 to 156%, (see the insets of Fig. 2(a), Fig. 2(b), and Fig. 1(c)). Recall that the ferro-coupling strength between the upper SyAF [Co/Pt] 3 layer and the Co 2 Fe 6 B 2 pinned layer directly affect the TMR ratio; i.e., a higher ferro-coupling strength leads to a higher TMR ratio 25 . Furthermore, H ex of the spin-valve increased slightly from 2.36 to 2.64 kOe, which would make the spin-valve more susceptible to read disturbance. Otherwise, when t W was increased from 0.30 to 0.48 nm, the M-H loop of the spin-valve changed abnormally because the thicker t W could not perfectly ferro-couple the upper SyAF [Co/Pt] 3 layer with the Co 2 Fe 6 B 2 pinned layer (see the red box in Fig. 2(c)). In particular, the squareness of the M-H loop of the Co 2 Fe 6 B 2 free layer drastically deteriorated, resulting in the TMR ratio falling from 156 to 0% (the insets of Fig. 2(b) and (c)). Note that the magnetic moment increased from 190 to 271 μemu, indicating that the upper SyAF [Co/Pt] 3 layer did not completely ferro-couple with the Co 2 Fe 6 B 2 pinned layer since the magnetic moment of the Co 2 Fe 6 B 2 free layer was only 190 μemu. In addition, H ex of the spin-valve greatly decreased from 2.64 to 1.16 kOe, which would cause a read failure.
In contrast, for the p-MTJ spin-valve using the single SyAF [Co/Pt] n layer, at a t W of 0.16 nm, its M-H loop was completely different from that of the spin-valve using the double SyAF [Co/Pt] n layer; i.e., there was no the squareness in the M-H loop for the single SyAF layer (see the red boxes in Fig. 2(a) and (d)). The spin-electron direction of the Co 2 Fe 6 B 2 pinned layer gradually rotated from upward to downward and saturated as the external perpendicular magnetic-field (H) was scanned from +6.5 to +1.5 kOe, (Fig. 2(d)). Then, the spin-electron direction of the Co 2 Fe 6 B 2 free layer rotated from upward to downward when H was scanned from + 0.2 to −0.2 kOe, meaning that the free layer showed good perpendicular magnetic characteristics and the TMR ratio was 173%. Finally, the spin-electron direction of the single SyAF [Co/Pt] 3 layer rotated from upward to downward when H was scanned from −0.2 to −6.5 kOe, resulting in an H ex of 2.52 kOe, which would be sufficient to avoid a read disturbance failure (Fig. 2(d)). In addition, when t W was changed from 0.16 to 0.30 nm, the squareness of the M-H loop of the Co 2 Fe 6 B 2 free layer slightly improved, resulting in an increase in the TMR ratio from 173 to 180% (compare the insets of Fig. 2(d), Fig. 2(e), and Fig. 1(c)). H ex of the p-MTJ spin-valve considerably increased from 2.52 to 3.44 kOe, probably improving the read disturbance of p-MTJ spin-valves. Furthermore, when t W changed from 0.30 to 0.48 nm, the M-H loop abnormally deteriorated because the thicker t W could not perfectly anti-ferro-couple the single SyAF [Co/Pt] 3 layer with the Co 2 Fe 6 B 2 pinned layer ( Fig. 2(f)). In particular, the squareness of the M-H loop of the Co 2 Fe 6 B 2 free layer greatly deteriorated, resulting the TMR ratio falling from 180 to ~0% (see the insets of Fig. 2(e) and (f)). In addition, H ex abruptly dropped from 2.64 to 1.16 kOe, which would probably cause a read disturbance failure. In summary, the M-H loop of the p-MTJ spin-valves using a single SyAF [Co/Pt] n layer had an H ex that was little higher than that of those using a double SyAF [Co/Pt] n layer, although the total thickness of the SyAF [Co/Pt] n layer was reduced considerably from 8.65 to 3.55 nm, which would not cause a read disturbance failure. However, the comparison of the M-H loops did not explain why the spin-valve using a single SyAF [Co/Pt] n layer had a higher TMR ratio than those using the double SyAF [Co/Pt] n layer. Thus, we investigated the face-centered-cubic (f.c.c) crystallinity of the p-MTJ spin-valves by using x-HRTEM. Here we should recall that the f.c.c crystallinity of the MgO tunneling barrier directly affects the probability of coherent tunneling of the spin electrons, determining dominantly the TMR ratio of a p-MTJ spin-valve 14,17-30 . Crystallinity of the MgO tunneling barrier. The f.c.c crystallinity of the MgO tunneling barrier was examined in the spin valves with the double SyAF [Co/Pt] n layer at t W = 0.30 nm (i.e., Fig. 2(b)) and a single SyAF [Co/Pt] n layer at t W = 0.24 nm (i.e., Fig. 2(e)), as shown in the x-HRTEM images of Fig. 3. The MgO tunneling barrier and capping layer of the spin-valve using the double SyAF [Co/Pt] n layer had a fluctuating surface like a sinusoidal wave (Fig. 3(a) and (c)). On the other hand, the MgO tunneling barrier and capping layer of the spin-valve using the single SyAF [Co/Pt] n layer had a flat surface ( Fig. 3(b) and (d)). The surface roughness of the tunneling barrier and capping layer became smoother as the SyAF [Co/Pt] n layer thickness was decreased from 8.65 nm (i.e., a double SyAF [Co/Pt] n layer) to 3.55 nm (i.e., a single SyAF [Co/Pt] n layer). In particular, the peak-to-valley of the MgO tunneling barrier (Δ P-V ) for the single SyAF [Co/Pt] n layer (i.e., 1.4 nm for t SyAF = 3.55 nm) was much less than that for the double SyAF [Co/Pt] n layer (i.e., 2.6 nm for t SyAF = 8.65 nm), as shown in Fig. 3(a) and (b).
Since the f.c.c crystallinity of the MgO tunneling barrier directly influences the coherent tunneling of spin-electrons in the spin-valves, x-HRTEM images of regions i and ii in Fig. 3(a) and (b) were taken (Fig. 3(c) and (d)). For the spin-valve with a double SyAF [Co/Pt] n layer, the fluctuating surface of the MgO tunneling barrier and capping layer (Fig. 3(c)) originated from the thicker SyAF [Co/Pt] n layer (t SyAF = 8.65 nm). In particular, the MgO capping layer showed almost amorphous regions (a and b in Fig. 3(c)) together with a region of locally f.c.c textured crystallinity (c in Fig. 3(c)) and it had a large thickness variation (i.e., ~0.64 nm at i and ~0.80 nm at ii in Fig. 3(c)). In contrast, the tunneling barrier looked almost completely f.c.c textured (d and e in Fig. 3(c)) and its thickness was ~0.96 nm at iii and 0.86 nm at iv in Fig. 3(c). Otherwise, the spin-valve with the single SyAF [Co/Pt] n layer had a much smoother MgO capping layer than that using the double SyAF [Co/Pt] n layer, (Fig. 3(c) and (d)) resulting in an uniform capping layer thickness (i.e., ~0.91 nm at i and ~0.85 nm at ii in Fig. 3(d)). In particular, except at the grain boundaries, the MgO capping layer was almost completely f.c.c, as shown in a and b in Fig. 3(d). Furthermore, the spin-valve with the single SyAF [Co/Pt] n layer had a tunneling layer with a flatter surface than that of the one with the double SyAF [Co/Pt] n layer (Fig. 3(c) and (d)). It had a uniform MgO tunneling barrier layer thickness (i.e., ~1.10 nm at iii and ~1.14 nm at iv in Fig. 3(d)). This result originated from the thickness difference between the single (t SyAF = 1.40 nm) and double SyAF [Co/Pt] n layer (t SyAF = 8.65 nm). The MgO tunneling barrier was f.c.c. except at the grain boundaries, as shown in c and d in Fig. 3(d). In particular, the tunneling barrier (i.e., 1.10~1.14 nm) for the spin-valve using the single SyAF [Co/Pt] n layer was quite thicker than that of the spin-valve with the double SyAF layer (i.e., 0.86~0.96 nm). These results indicate that the f.c.c. crystallinity of the MgO tunneling-barrier layer for the spin-valve using single SyAF [Co/Pt] n layer was much superior to that of the spin-valve using the double SyAF [Co/Pt] n layer. Here, we should recall that larger surface roughness in the MgO tunneling barrier reduces the hybridization of the Co-O and Fe-O atoms at the interface between the Co 2 Fe 6 B 2 free layer and tunneling barrier which degrades the i-PMA characteristics of the p-MTJ 20,31,32 . In addition, better f.c.c crystallinity of the MgO tunneling barrier leads to a higher coherent tunneling ability of spin-electrons. Thus, both surface roughness and f.c.c crystallinity directly and dominantly affect the TMR ratio of the p-MTJ spin-valves; i.e., smaller surface roughness and better f.c.c crystallinity of the MgO tunneling barrier leads to a higher TMR ratio 14,[17][18][19][20][21][22][23][24][25][26][27][28][29][30] . Thus, as revealed in the x-HRTEM image, the MgO tunneling barrier of the spin-valve using the single SyAF [Co/Pt] n layer obviously showed a smaller surface roughness and better f.c.c crystallinity compared with the barrier of the spin-valve with the double SyAF [Co/Pt] n layer. As a result, the p-MTJ spin-valve using the single SyAF [Co/Pt] n layer (~180%) achieved a higher TMR ratio than that of the one using the double SyAF [Co/Pt] n layer (~156%).

Discussion
Double MgO-based p-MTJ spin-valve with a top Co 2 Fe 6 B 2 free layer using a double SyAF [Co/Pt] n layer ( Fig. 1(a)) have been intensively studied in order to achieve TMR ratios higher than 150% in order to ensure a memory margin and enough thermal stability (Δ) for a ten-year retention-time at BEOL temperatures higher than 350 °C. However, the high surface roughness of the MgO tunneling barrier that originates from the thick double SyAF [Co/Pt] n layer (i.e., 8.65 nm) potentially limits further enhancement of the TMR ratio. As a solution, a double MgO-based p-MTJ spin-valve structure with a top Co 2 Fe 6 B 2 free layer using a single SyAF [Co/Pt] n layer ( Fig. 1(b)) was developed that has a thinner single SyAF [Co/Pt] n layer (i.e., 3.55 nm) and a buffer layer bridging the single SyAF [Co/Pt] n layer and the Co 2 Fe 6 B 2 pinned layer (i.e., Co/Pt/Co layer). This device structure demonstrated a sufficient anisotropy exchange field (i.e. 3.44 kOe) for avoiding read disturbance failures and a TMR ratio about 25% higher than that of a p-MTJ spin-valve using double SyAF [Co/Pt] n layer. In particular, the f.c.c crystallinity of the MgO tunneling barrier was improved by the greatly reduced surface roughness of the MgO tunneling barrier (i.e., 1.4 nm). Therefore, the double MgO-based p-MTJ spin-valve with a top Co 2 Fe 6 B 2 free layer using a single SyAF layer would be a promising spin-valve structure for future device application such as a terra-bit integration stand-alone memory, an embedded memory, a storage class memory, etc.

Methods
All double MgO-based p-MTJ spin-valves were fabricated using a 12-inch-wafer multi-chamber cluster-magnetron sputtering-system under a high vacuum of less than 1 × 10 −8 torr. In particularly, the double MgO-based p-MTJ spin-valve structure with a top Co 2 Fe 6 B 2 free layer using a double SyAF [Co/Pt] n layer were fabricated as a vertical stack containing a 12-inch SiO 2  MgO-based p-MTJ spin-valve with a top Co 2 Fe 6 B 2 free layer using a single SyAF [Co/Pt] n layer were fabricated wherein the ratio of the number of [Co/Pt] multi-layers between the upper and lower SyAF [Co/Pt] n layer was varied from 3:6 (i.e., a double SyAF [Co/Pt] n layer) to 0:3, as shown in Fig. 2 (b). In addition, the bridging buffer layer was a Co/Pt/Co layer instead of the single Co layer (compare Fig. 1(a) and (b)). Note that the bridging buffer layer is necessary to ferro-couple the Co 2 Fe 6 B 2 free layer with the SyAF [Co/Pt] n layer. The spin-valves were ex-situ annealed at 350 °C for 30 min under a vacuum below 10 −6 torr and a perpendicular magnetic field of 3 tesla. The TMR ratios of the 12-inch wafer p-MTJ spin-valves were estimated by using current in-plane tunneling (CIPT) at room temperature. Afterward, the wafers were cut into 1 × 1 cm 2 pieces. The magnetic properties (out-of-plane and in-plane) for the two different spin-valves were characterized by using vibrating sampling magnetometer (VSM) at room temperature. The crystallinity was estimated by using x-HRTEM at an acceleration voltage of 200 keV.