Abstract
We investigated the structural and magneto-transport properties of Fe/SrO/MgO/Fe magnetic tunnel junctions (MTJs), in which SrO has a large lattice mismatch (Δa/a ∼ 21%) with Fe. Despite such a large Δa/a, structural analyses revealed that a fully epitaxial Fe(001)/SrO(001)/MgO(001)/Fe(001) structure was successfully formed. We observed high magnetoresistance (MR) ratios up to 98% at 20 K (65% at room temperature), suggesting that coherent spin-polarized tunneling occurs through the SrO(001) tunnel barrier. We also demonstrated that the resistance-area product and bias-voltage (Vhalf), at which the MR ratio reaches half of the zero-bias value, were comparable to the values reported in epitaxial Fe/MgO/Fe MTJs. Since SrO has a smaller Δa/a with Si (∼5%) than that between MgO and Si (∼23%), the results offer that this material is a promising candidate as a high-quality tunnel barrier for Si-based spin transport devices.
Export citation and abstract BibTeX RIS
1. Introduction
Epitaxial tunnel barrier is the core component of magnetic tunnel junctions (MTJs) showing giant tunneling magnetoresistance (MR) effect at room temperature (RT) based on the phenomenon called coherent spin-polarized tunneling.1–12) As examples of the epitaxial tunnel barrier, MgO,1–4) MgAl2O4,5,6) GaAl2O4,7) ZnO,8–10) and GaOx11,12) have been reported so far. These insulators are of rock salt- or spinel-type crystal structures having small lattice mismatches (Δa/a) with conventional 3d-ferromagnetic (FM) of bcc-Fe(Co). This characteristic enables to grow fully epitaxial MTJs that exhibit the giant tunneling MR effect, which is suitable for practical use. On the other hand, however, there are limitations concerning the range of the variations of lattice constants that can be controlled artificially for the epitaxial tunnel barrier. To broaden the application fields of the epitaxial FM contact, it is greatly beneficial to develop novel type of epitaxial tunnel barriers having various lattice constants.
Fabrication of an epitaxial FM tunnel contact on a Si channel is the key to developing semiconductor-based spin-transport devices such as a spin metal-oxide-semiconductor field-effect transistor (spin-MOSFET). To date, MgO has been exclusively used as an epitaxial tunnel barrier on Si.13–18) We recently have succeeded in injecting highly spin-polarized current (P ∼ 0.9 at 10 K; P is the spin polarization of tunneling electrons) from epitaxial Fe(001)/MgO(001) tunnel contacts into Si(001) channel.17,18) Such high P gives a potential MR ratio of several hundred %, according to the standard spin-transport model for a lateral device consisting of semiconductor channel with two FM contacts, like a spin-MOSFET.19) However, such high P has only been achieved in a very high resistance-area product (RA) region (>1 MΩ μm2) for Fe/MgO contact,17,18) which significantly suppresses the MR ratio.19) In the low RA region (<10 kΩ μm2), on the other hand, the tunneling electrons can no longer keep such high P (for example, P < 25% at 10 K).17,18) The observed degradation of P in the low RA region can possibly be attributed to a poor crystalline quality of the very thin (<1 nm) MgO tunnel barrier caused by the large Δa/a between MgO and Si (22.6%, see Table I). Consequently, the MR ratio so far achieved in Si-based lateral devices has been less than 1%.13) Therefore, it is desirable to introduce a new tunnel barrier material which, from the standpoint of high-quality thin epitaxial tunnel barrier, can be adequately lattice-matched with Si, and thereby achieving high MR effect in Si-based lateral device.
Table I. Lattice constants (a) and lattice mismatches of oxides (MgO and SrO) with Si (Δa/aSi) and Fe (Δa/aFe). The Δa is defined as aX − aY (X = oxide, and Y = Si or Fe), aSi and aFe are respectively 0.543 nm and 0.287 nm. The Δa/aFe are obtained by considering 45° rotation of the unit cell of oxides with regard to the Fe unit cell.
| Oxide | a (nm) | Δa/aSi (%) | Δa/aFe (%) |
|---|---|---|---|
| MgO | 0.420 | 22.6 | −3.55 |
| SrO | 0.516 | 4.97 | −21.4 (−21.6a)) |
a)Estimated from the high resolution HAASF-STEM image in the present study.
SrO (band gap of 5.8 eV20)) is possibly suitable for such tunnel barrier because of its relatively small Δa/a with Si (4.97%) having the same crystal structure as MgO (rock salt-type). In addition, direct growth of a high-quality epitaxial of SrO(001) film on Si(001) has been demonstrated from an initial stage of the growth.21) The layer-by-layer growth of the SrO(001) on Si(001) has been observed up to the SrO thickness of ∼1.5 nm, which is sufficient in terms of thickness for the tunnel barrier.21) Despite those promising features, application of SrO to a spin-transport device has been limited to an amorphous tunnel barrier on graphene.22) Besides, fundamental physical constants of tunneling, such as barrier height (ϕ) of epitaxial SrO, have not been reported yet. In addition, the question remains whether coherent spin-polarized tunneling can be achieved through epitaxial SrO(001) tunnel barrier.
This study reports the fabrication of fully epitaxial Fe(001)/SrO(001)/MgO(001)/Fe(001) MTJs, where MgO tunnel barrier acts as the underlying layer for the epitaxial growth of SrO tunnel barrier. High MR ratios up to 98% at 20 K (65% at RT) were observed, indicating the existence of coherent spin-polarized tunneling through the SrO tunnel barrier. The results suggest that SrO is a promising tunnel barrier for Si-based spintronic devices.
2. Experimental procedures
MTJ films as presented in Fig. 1 were grown by molecular beam epitaxy (MBE). The films consisted of a Au cap (10 nm)/Co pinned layer (20 nm)/Fe top electrode (10 nm)/SrO tunnel barrier (1.4 nm)/MgO tunnel barrier (0.8 nm)/Fe bottom electrode (30 nm)/MgO buffer layer (5 nm) on a MgO(001) substrate. The MgO tunnel barrier enables the SrO tunnel barrier to grow epitaxially.23) The source materials were evaporated using electron-beam guns (for Fe, SrO, and MgO) and Knudsen-cells (for Au and Co). Single-crystal SrO granules and MgO block were used as source materials. Prior to the growth, the MgO substrates were cleaned by an ultrasonic cleaner with acetone and isopropanol, and thermally annealed at 800 °C for 10 min in the MBE chamber with a base pressure 2 × 10−9 Torr. The MgO buffer layer and Fe bottom electrode were deposited on the substrate at 100 °C, followed by an in situ annealing at 300 °C for 10 min to improve the surface morphology of the Fe bottom electrode. Then, the MgO tunnel barrier was grown on the Fe bottom electrode at RT. Subsequently, the SrO tunnel barrier was deposited on the MgO tunnel barrier at RT under an O2 pressure of 1–3 × 10−7 Torr. Growth rates of the MgO and SrO tunnel barriers were 0.02–0.03 nm s−1. The Fe upper electrode was grown on the SrO tunnel barrier at RT, and then annealed for 10 min at 300 °C to reduce the dislocation density at the Fe/SrO or SrO/MgO interfaces. Finally, Co-pinned and Au-cap layers were deposited onto the Fe top electrode at RT. As a reference, the same MTJ stack without the MgO tunnel barrier was also grown.
Fig. 1. Schematic structure of MTJ stack and in situ annealing temperature adopted during the deposition processes.
Download figure:
Standard image High-resolution imageStructural properties of the MTJ films were investigated by in situ reflection high energy electron diffraction (RHEED), ex situ cross-sectional high-angle annular dark field scanning transmission electron microscope (HAADF-STEM), and the elemental mapping by energy-dispersive X-ray spectroscopy (EDX). For the magneto-transport measurements, the films were patterned into tunnel junctions with active areas from 3 × 12 μm2 to 6 × 24 μm2 using conventional micro-fabrication techniques (e.g. photolithography, Ar ion milling, and SiO2 sputtering). The measurements were carried out using a conventional DC two-probe method. The magnetic fields were applied parallel to the major axis of the junction corresponding to the easy axis of the magnetization direction of the Fe electrodes.
3. Results and discussions
3.1. Structural analysis
Figure 2 shows the RHEED patterns of (a) the Fe bottom electrode, (b) the MgO tunnel barrier, (c) the SrO tunnel barrier, and (d) the Fe top electrode, respectively (along the [100] azimuth of the MgO substrate). The RHEED images of the Fe bottom electrode [Fig. 2(a)] and MgO tunnel barrier [Fig. 2(b)] showed streaky patterns, indicating single-crystalline Fe(001) and MgO(001), respectively. It was confirmed that the in-plane crystal orientation between the MgO tunnel barrier and Fe bottom electrode was MgO[100] ∣∣ Fe[110], i.e. the unit cell of the MgO was turned 45° with regard to the Fe bottom electrode.4,9,11,12) For the SrO tunnel barrier, we observed elongated spotty patterns having a four-fold symmetry [Fig. 2(c)], indicating the formation of single-crystalline SrO(001) tunnel barrier. The patterns of the Fe top electrode as displayed in Fig. 2(d) were identical to those of the Fe(001) bottom electrode [see Fig. 2(a)]. As a result, we can say that a fully epitaxial Fe(001)/SrO(001)/MgO(001)/Fe(001) structure was successfully formed. For the reference sample without the MgO tunnel barrier, we did not obtain an epitaxial SrO tunnel barrier, and therefore epitaxial Fe top electrode could not be formed.
Fig. 2. RHEED patterns of the (a) Fe bottom electrode after in situ annealing, (b) MgO tunnel barrier, (c) SrO tunnel barrier and (d) Fe top electrode after in situ annealing, respectively ([100] azimuth of MgO substrate).
Download figure:
Standard image High-resolution imageFurther structural analyses were performed on the fully epitaxial film by means of HAADF-STEM and EDX observations as illustrated in Figs. 3(a)–3(c). These images revealed excellent uniformities of the SrO and MgO tunnel barriers and steep interfaces without interdiffusion among each layer. Note that, from the high resolution HAADF-STEM image [Fig. 3(b)], misfit dislocations in every several atoms at the Fe/SrO and SrO/MgO interfaces (corresponding to the line density of ∼1 × 107 cm−1) were clearly observed. By assuming a 45° in-plane rotation between the Fe and SrO layers and a cube-on-cube relation between the SrO and MgO tunnel barriers, the in-plane Δa/a between the SrO tunnel barrier and Fe top electrode, and that between the SrO tunnel barrier and bottom electrode are both estimated to be −21.6%. The estimated value is very close to the expected value from the bulk material (−21.4%, see Table I). Accordingly, the in-plane crystal orientations were determined as top Fe[110] ∣∣ SrO[001] ∣∣ MgO[001] ∣∣ bottom Fe[110], respectively. Note that the 45° in-plane rotation between the MgO(001) tunnel barrier and Fe(001) electrodes is a preferable crystal orientation to achieve coherent spin-polarized tunneling.1–3) Since SrO has the same crystal structure as MgO, the result implies that we can expect coherent tunneling with this type of MTJ.
Fig. 3. (Color online) (a) Low and (b) high resolution cross-sectional HAADF-STEM images, and (c) EDX elemental mappings of Fe, O, Mg, and Sr of the Fe/SrO/MgO/Fe stack ([100] azimuth of MgO substrate). Dislocations in the HAADF-STEM image are indicated by T shape symbols. Dashed lines on the EDX mappings indicate the barrier-electrode interfaces, for guides to the eye.
Download figure:
Standard image High-resolution imageNote that it may be difficult to largely reduce the number of the dislocations at the Fe/SrO and SrO/MgO interfaces by process engineering since the observed dislocations are inevitably introduced for a heteroepitaxial growth with large Δa/a.
3.2. Magneto-transport properties
Typical MR curves of the epitaxial MTJ are shown in Fig. 4(a). Here, the MR ratio is defined as (RAP–RP)/RP, where RP and RAP are the junction resistances in parallel (P) and antiparallel (AP) magnetization states, respectively. We observed MR ratios up to 98% at 20 K and 65% at RT, respectively. The observed MR ratios are almost twice as high as those reported in a polycrystalline Fe/amorphous GaOx/MgO(001)/Fe(001) MTJ (50% at 20 K and 34% at RT, respectively),11) where the coherent spin-polarized tunneling is considerably suppressed due to the polycrystalline nature of the Fe top electrode and amorphous GaOx tunnel barrier. This strongly suggests that the Fe(001)/SrO(001) including SrO(001)/MgO(001) interface is capable of realizing coherent spin-polarized tunneling in the same manner as the MgO(001)/Fe(001), and Δ1 state acts as the major tunneling channel in the P state.1–4) In addition, it was found that the RP has small temperature dependence, namely, relative changes in the RP between 20 K and RT were very small (2%–4%) compared with those for the RAP (10%–13%). This is a typical feature of fully epitaxial MgO- and MgAl2O4-based MTJs,4,5) in which coherent tunneling has been both experimentally and theoretically demonstrated. The result supports our conclusion that the coherent tunneling is achieved in the epitaxial Fe/SrO/MgO/Fe MTJ. Also, the result suggests an excellent thermal stability in this MTJ since the in situ annealing up to 300 °C has been conducted.
Fig. 4. (Color online) (a) Magnetoresistance curves of the Fe/SrO/MgO/Fe MTJ at 20 K and RT applying 10 mV voltage, and (b) bias-voltage (V) dependence of the normalized MR at RT. Resistance-area product (RA) in the parallel (P) magnetization state were 10–20 kΩ μm2 at RT.
Download figure:
Standard image High-resolution imageNote that we did not observe any MR effect nor a nonlinear (tunnel-like) current–voltage characteristics in the reference MTJ, implying that there are many imperfections such as pinholes in the polycrystalline SrO tunnel barrier.
It is interesting to know how the asymmetric barrier/electrode structure with such a high density of dislocations at the interfaces affects the spin-dependent tunneling. In general, bias-voltage (V) dependence of the MR ratio is sensitive to the barrier and barrier/electrode interface qualities. Namely, poor barrier and/or barrier/electrode interface qualities result in a low bias-V at which the MR ratio reaches half of the zero-bias value (Vhalf). As a result, rapid decrease in the MR ratio often occurs in a bias-V direction where the electrons tunnel into the electrode having a lower interface quality.24,25) In Fig. 4(b), the MR ratio at RT is plotted as a function of bias-V for the epitaxial Fe/SrO/MgO/Fe MTJ. Here, positive bias is defined as the bias direction where the electrons tunnel from the Fe bottom electrode (MgO/Fe interface) into the Fe top electrode (Fe/SrO interface). The plot was almost symmetric despite the asymmetric structure, and the Vhalf became as high as 800 mV regardless of the bias directions. The observed Vhalf is higher than typical values for the MTJs with an amorphous AlOx tunnel barrier (range from 300 to 600 mV) and even comparable to those in lattice-matched systems of epitaxial Fe/MgO/Fe,3) Fe/MgAl2O4/Fe5) and Fe/GaOx/MgO/Fe11) MTJs. Consequently, we could not observe clear evidence of the influence on the MR ratio for the asymmetric barrier/electrode structure with many dislocations at the interfaces. Conversely however, the observed high Vhalf in both bias directions are favorable for efficient spin-injection/detection schemes in a Si-based lateral device with two FM contacts.
3.3. Estimation of the physical constant of tunneling for SrO(001) tunnel barrier
We estimated the physical constants of the SrO(001) tunnel barrier from simulating the current density versus voltage (J–V) characteristics of the epitaxial Fe/SrO/MgO/Fe MTJ based on the calculation method developed by Grover and Moddel.26) For simplicity, only a tunneling channel of the Δ1 state in the majority-spin band was considered; therefore, the simulation was conducted in the P state.9,27) Other possible transport processes, such as thermionic emission, hopping conduction, and inelastic tunneling, were ignored. The J–V characteristics were then determined by the work function of Fe(001) electrodes (4.6 eV28)), the electron affinity (χ), relative dielectric constant (εr), and effective mass of tunneling electron (m*), and thickness of the MgO(001) and SrO(001) layers. Since the χ, εr and m* of MgO(001) thin layer have been already deduced,9,29) the simulation has been performed using the χ, εr and m* of the SrO(001) as variables. Nominal thicknesses of the SrO (1.4 nm) and MgO (0.8 nm) layers were employed.
Figure 5 shows the experimental (solid line) and simulated (red circles) J–V characteristics of the epitaxial MTJ in the P state at RT. The experimental data is well reproduced by the simulation, and the tunneling parameters determined from the simulations are summarized in Table II, together with those obtained for MgO(001) tunnel barrier in the past.9,29) Note that all the parameters of the SrO(001) are close to those of the MgO(001), which is consistent with the fact that the present MTJ has a comparable RA (10–20 kΩ μm2 in the P state at RT) to that of the epitaxial Fe/MgO/Fe MTJ with an equivalent barrier thickness of 2.2 nm.4)
Fig. 5. (Color online) Experimental (solid line) and simulated (red circles) current density versus voltage (J–V) characteristics of the Fe/SrO/MgO/Fe MTJ in the P state at RT.
Download figure:
Standard image High-resolution image
Table II. Electron affinity (χ), relative dielectric constant (εr), and effective mass (m*) of SrO(001) tunnel barrier determined by simulating J–V characteristics in the P state at RT, together with those obtained for MgO(001) in the past.9,29) Barrier heights (ϕ) correspond to energy difference between χ of the oxides and work function of Fe(001) of 4.6 eV.28)
| Oxides | χ (eV) | εr | m* | ϕ (eV) |
|---|---|---|---|---|
| SrO | 0.40 | 7.0 | 0.13 | 4.2 |
| MgO | 0.65a) | 8.8b) | 0.10a) | 4.0a) |
4. Conclusions
We studied the structural and magneto-transport properties of the lattice-mismatched Fe/SrO/MgO/Fe MTJ. The RHEED and STEM observations revealed a fully epitaxial Fe(001)/SrO(001)/MgO(001)/Fe(001) structure having many misfit dislocations at the Fe/SrO and SrO/MgO interfaces. Despite the existence of such misfit dislocations, high MR ratio up to 98% at 20 K (65% at RT) was observed, indicating that coherent spin-polarized tunneling takes place through the SrO(001) tunnel barrier. The Vhalf values in both bias directions were as high as 800 mV, which are comparable to the reported value in the epitaxial Fe/MgO/Fe MTJs. We found that all the tunneling parameters such as ϕ and m* of the SrO(001), estimated from the J–V characteristics, are close to those of the MgO(001), resulting from the similar RA to that of the epitaxial Fe/MgO/Fe MTJ. Since SrO has a much smaller Δa/a with Si, SrO(001) appears as a promising epitaxial tunnel barrier for achieving high MR ratio in Si-based lateral spin transport devices.
Acknowledgments
This work was supported by JSPS KAKENHI (Grant No. 18K13807, A.S.).