The  Magnetotransport  properties  in  perpendicularly  magnetized  tunnel  junctions  using  an

electrode

4.2.1. Experimental procedures

Stacked MgO(001) substrate/MgO (5)/Cr (30)/Fe (0.7)/MgO barrier (1.8) (unit in nm) structures were fabricated using an ultra-high vacuum electron beam evaporation system with a base pressure below 1×10⁻⁸ Pa. The surface structure of each layer was observed using reflection high-energy electron diffraction (RHEED). The epitaxial growth from the substrate to the MgO barrier was determined as MgO(001)[100]//Cr(001)[110]//Fe(001)[110]//MgO(001)[100] from the RHEED patterns. The Cr buffer layers were deposited at RT and post-annealed at TCr = 800 and 700°C in order to obtain different interface PMA at the Fe/MgO interface. According to our previous work, the interface PMA at the Fe/MgO interface was strongly dependent on the post-annealing temperature of the Cr buffer layer, which resulted in different surface crystalline quality of the Cr buffer, as well as Fe layer.¹⁰ The 0.7 nm thick Fe layer was deposited at 150°C on the Cr buffer layer, and subsequently post-annealing at 250°C was carried out to make the Fe layer flat. It was confirmed by transmission electron microscopy observation¹⁰ that an Fe layer of mostly 5 monolayers (MLs) in thickness was obtained in this process. After the MgO-barrier layer deposition, the samples were post-annealed at 400°C in the same chamber. The samples were then transferred to an rf magnetron sputtering system with a base pressure below 3×10⁻⁷ Pa and structures of Co20Fe60B20 (tCoFeB = 1.2–1.4)/Ta (4.5)/Ru (15) were deposited on the MgO barrier using a linear motion shutter, as depicted in figure 1. In each sample, we also prepared the area where no CoFeB top electrode is deposited, see Figure 4.14, to investigate the interface PMA characteristic at the Fe/MgO interface.

Magnetization (M)–magnetic field (H) loops were measured at RT using a vibrating sample magnetometer, and the magnetotransport properties of the unpatterned MTJs were characterized at RT through current-in-plane tunneling (CIPT) measurement after ex-situ annealing for 30 min at temperatures (Tann) ranging from 250 to 450°C. After annealing at 450°C, the samples were patterned into the MTJ pillars with an active area of 10 × 5 µm² through conventional UV lithography combined with a lift-off technique and Ar ion etching. The patterned MTJs were characterized through a dc four-probe method at RT.²⁵

4.2.2. Results and Discussion

First of all, we measured magnetic anisotropy characteristics for the Cr/Fe/MgO/Ta/Ru structures, whose Cr buffer layers were annealed at different temperatures, T_Cr = 800 (Series-I) and 700°C (Series-II). In Figure 4.15 (a) and (b), in-plane and out-of-plane M–H loops for the films of Series-I and Series-II are shown, respectively.

  ⁹⁵

Figure 4.14 Schematic illustration of the MTJ stacked structure. [25]

Figure 4.15 M–H loops for two Cr(30)/Fe(0.7)/MgO(1.8)/Ta(4.5)/Ru(15) (in nm) stacks with annealing temperatures for Cr layers, T_Cr, (a) T_Cr = 800°C (Series-I) and (b) T_Cr = 700°C (Series-II). The whole stacks were post-annealed at Tann = 400°C. [25]

The Ki was determined using the simple relationship Ki = (Keff − KV) × tFe, where Keff is the effective PMA energy density and K_V is the volume anisotropy energy density, which can be simply treated as a shape anisotropy energy density (−µ0MS2/2, where MS is the saturation magnetization), and tFe is the thickness of the Fe layer. Here, the Keff is calculated from the area enclosed by the in-plane and out-of-plane

  ⁹⁶

magnetization curves and the y-axis. Although the two samples have almost the same MS, they exhibit significantly different magnetic anisotropy characteristics. (The Fe dead layer was estimated to be less than 0.04 nm in thickness.) In the case of Series-I, it is clearly seen that the easy-axis is perpendicular to the film plane with a K_i of ~ 1.5 mJ/m². On the other hand, the K_i of Series-II was estimated to be smaller than 1.0 mJ/m². The contribution of the interface magnetic anisotropy at Cr/Fe interfaces on this interface PMA was negligibly small which was confirmed by using a Cr/Fe(0.7 nm)/Cr structure. This large difference in interface PMA of the two samples is consistent with our previous study, which showed the large difference in interface PMA of Fe/MgO bilayers resulted from different annealing temperatures of Cr buffers¹⁰.

Figure 4.16 (a) TMR vs. out-of-plane H curves for the MTJs in Series-I and Series-II. The inset is the M–H loop for the unpatterned Series-I after annealing at 450°C, and (b) TMR ratios as a function of the T_ann with respect to each tCoFeB of the two series (Series-I : solid lines, Series-II: dashed lines). [25]

Figure 4.16 (a) shows the TMR–perpendicular H curves for the Series–I and –II MTJs with tCoFeB = 1.4 nm and Tann = 450°C, characterized using CIPT for unpatterned films. The curves for both samples clearly show high and low resistance states. This shows that both the bottom Fe and the top CoFeB electrodes are perpendicularly magnetized and stable parallel (P) and antiparallel (AP) configurations were achieved.

However, the switching field of the bottom electrode (Fe) for each sample is quite different, owing to the difference in the coercivity, i.e., difference in the PMA characteristics. The maximum TMR ratio of up to 95% was achieved for Series-I, whereas that of 30% was obtained for Series-II. Here, the TMR ratio is defined as (RAP – RP)/RP, where RP and RAP denote the tunneling resistances for P and AP configurations, respectively. Figure 4.16 (b) shows the TMR ratios of the p-MTJs as a function of Tann with tCoFeB = 1.3 and 1.4 nm. (In the case of tCoFeB = 1.2 nm, stable P and AP states were not obtained for both samples.)

  ⁹⁷

Independent of the CoFeB electrode thickness, the TMR ratio of Series-I exhibiting a larger PMA for the bottom Fe electrode was always larger than that of Series-II for Tann ranging from 250 to 450°C. This difference in the TMR ratios between Series–I and –II is considered to be the consequence of two different annealing temperatures of the Cr buffer layers, same as the differences in the K_i values for the bottom Fe layer, because the top CoFeB layers have identical structures. In other words, despite the different physical origin between the interface PMA and the TMR effect, both phenomena are extremely sensitive to a small change in the surface crystalline quality of the Cr buffer layer, in this system. It is also to be noted that the shape and amplitude of the TMR curve for the patterned MTJ was almost identical to that for the unpatterned film.

From a more technical point of view, a large Ki for the bottom Fe electrode is favorable for achieving stable P and AP magnetization states. It is also worth noting that the large TMR ratio showed no severe decrease even at the higher annealing temperature of Tann = 450°C. This implies that the interface PMA of the Fe/MgO/CoFeB p-MTJ structure exhibits a better endurance under such a high annealing temperature compared to the other systems. It was reported that the TMR ratio in the p-MTJs with the CoFeB/MgO/CoFeB structure is likely to steeply decrease after annealing at over ~350°C mainly due to a deterioration in the PMA characteristics of the CoFeB layers^26,27. This is in contrast to our case exhibiting a high thermal endurance of the PMA, as shown in the TMR–H loops (inset of Figure 4.16 (a)).

Figure 4.17 dI/dV curve measured at RT for the MTJ in Series-I with tCoFeB = 1.4 nm after annealing at Tann = 450°C. [25]

In order to investigate the magnetotransport characteristic in more detail, we evaluated the differential conductance (dI/dV) of the patterned p-MTJ. In Figure 4.17, the dI/dV curve for the p-MTJ with tCoFeB = 1.4 nm and Tann = 450°C for the P configuration at RT is shown. The negative bias voltage indicates that the electrons flow from the top CoFeB to the bottom Fe. A dI/dV curve generally reflects the density of states of the electrodes. Interestingly, it was found that, besides an asymmetric feature, a clear peak was observed at around V = 0 V in the dI/dV curve. From the previous studies on SDRT in Fe/MgO-based MTJs^28,29, the

  ⁹⁸

peaks at around 0 V and −0.6 V in the curve are attributed to the SDRT process through the QW states in the ultrathin Fe(001). These QW states are limited to Δ1 symmetry because the ultrathin Fe(001) is sandwiched between the MgO(001) insulating barrier and the Cr(001) metallic barrier^24,28,29. The tunneling process in this p-MTJ occurs in the asymptotic regime of the MgO barrier, i.e. the Bloch states having non-zero in-plane momentum (k|| ≠ 0) and having Δ2, Δ2’, or Δ5 symmetries are attenuated by the filtering effect, so that Δ1

symmetry electrons governs the tunneling process. Furthermore, the SDRT effect through the QW states confined in the ultrathin Fe(001) layer indicates that the spin, symmetry, and wave vector of electrons are conserved through the coherent tunneling process. It is noted that the appearance of a strong peak at 0 V shows that the dominant Fe thickness of the MTJ is five MLs, ~0.7 nm, while a relatively weak peak at around −0.6 V indicates that the local existence of four or six MLs in the MTJ area.^28,29 Here, we can conclude that the PMA at the Fe/MgO interface is likely to little influence the Δ₁ electrons’ transport. The present results may give an insight into the understanding of the coherent and the resonant tunneling under a perpendicular magnetization configuration. For further understanding, characterization of the SDRT effect for the MTJ structures comprising the Fe layers with various and more precisely controlled thicknesses is needed.