Guidelines for attaining optimal soft magnetic properties in FeAlSi films Open Access

Sendust alloy (Fe_73.7Al_9.7Si_16.6atm %; hereinafter “FeAlSi”) is a soft magnetic material invented in 1937 by Masumoto and Yamamoto.¹ FeAlSi has an ideal soft magnetic property comparable to that of NiFe,² a conventional soft magnetic material. By fine-tuning the composition of FeAlSi alloys, Masumoto and Yamamoto discovered a specific composition (Sendust central composition, Fe_73.7Al_9.7Si_16.6atm%) with very high magnetic permeability by preparing numerous samples with different compositions. The soft magnetic property of the Sendust central composition stems from the magnetocrystalline anisotropy constant K₁, and the saturation magnetostriction constant λ_s becomes zero simultaneously.^3–6 The crystal structure of FeAlSi is a D0₃-ordered structure, which is an ordered structure of bcc(A2).⁷ As the degree of D0₃ order decreases, it changes into a B2⁷ or A2 structure and causes the soft magnetic property to degrade.⁸ FeAlSi has various applications (e.g., magnetic heads using bulk FeAlSi^8–10), and the compatibility of the FeAlSi central composition and the D0₃-ordered structure is essential for bringing out its soft magnetic properties. Previous studies on μm-order FeAlSi films using sputtering and evaporation and prepared films revealed soft magnetic properties and D0₃-ordered structures similar to that of bulk.^11–14 Although there have been subsequent studies on FeAlSi, none have examined FeAlSi films with nm-order thickness. Although high-quality sputtering techniques have been established and the thin-film properties of various materials have been investigated in detail, research on FeAlSi thin films has been relatively insufficient. This should be addressed because the bcc soft magnetic material, FeAlSi, may be suitable for application to tunnel magneto-resistive (TMR) sensors, which have been attracting attention in the field of spintronics.¹⁵ TMR sensors consist of three thin-film layers: a pinned layer, an insulator, and a free layer. The sensitivity of the TMR sensor is expressed as TMR ratio/2H_k,¹⁵ where H_k is the magnetic anisotropy field of the free layer. The bcc structure is expected to have a high TMR ratio due to the coherent tunneling of the Δ₁ electron,¹⁶ which makes FeAlSi a promising free layer material for attaining both a high TMR ratio and a low H_k. As such, it is crucial to investigate the magnetic properties of FeAlSi thin films in detail. In our previous work,¹⁷ we fabricated nm-order FeAlSi polycrystalline sputtered films for application to TMR sensors and found that the magnetic tunnel junctions (MTJs) with FeAlSi electrodes exhibited a TMR effect with ideal soft magnetic properties. We also fabricated FeAlSi epitaxial films, but their soft magnetic properties were inferior to that of bulk due to insufficient composition adjustment in sputtered FeAlSi films.¹⁸ The purpose of our present study is to obtain soft magnetic properties comparable to those of bulk by finely controlling the composition and atomic ordering of the FeAlSi films.

All films were deposited on MgO[001] substrates by DC/RF magnetron sputtering (P_base < 2 × 10⁻⁶Pa). The stacking structure was MgO-sub/MgO(20)/FeAlSi(30)/Ta(5) (thickness in nm). The calculated lattice mismatch between FeAlSi and the MgO buffer layer was a relatively low value of 4.7%, so epitaxial growth of FeAlSi films on MgO was expected. The structure of the FeAlSi layers was measured with x-ray diffraction (XRD) of out of the plane 2θ–θ scan, (111)-plane ϕ scan, and in-plane reciprocal space mapping (RSM). The magnetic properties were investigated using a vibrating sample magnetometer (VSM), ferromagnetic resonance (FMR), and the optical lever method.¹⁹ FMR was measured to determine the exact magnetocrystalline anisotropy constant K₁. The rotating magnetic field was applied in the in-plane direction of the film, and K₁ and K_u were calculated from the angular dependence of the resonant field by the following equation:

where K₁ is a cubic magnetic anisotropy constant, K_u is a uniaxial magnetic anisotropy constant, ω is the resonance frequency, and γ is a gyromagnetic ratio constant. K₁ and K_u represent the fourfold and twofold in-plane components of the magnetocrystalline anisotropy constant, respectively. The optical lever method was used to determine the magnetostriction constant of the FeAlSi films. We applied a rotating magnetic field in the in-plane direction of the films and observed the change in the reflection position of the laser beam due to the distortion of the sample caused by the inverse magnetostriction effect. The shift between the sample distortion and the laser reflection position is expressed by the following equation:

where $δl/l$ is the film distortion, $dlaser$ is the laser reflection position, $E$ is Young's modulus, $t$ is the thickness, $p$ is the spot position, and $L$ is the sensor position.

Film strain was measured along the FeAlSi [110] direction, and then $λ111$ was calculated using the following equation:²⁰ 

where $aϕ$ is an offset term.

All films were annealed at T_FeAlSi = 300–600 °C for 1 h without applying a magnetic field to improve the degree of D0₃ order and soft magnetic properties. As mentioned above, the soft magnetic properties of FeAlSi are very sensitive to the composition of the film, but the composition of sputtered films generally deviates from the target composition.²¹ In this study, five samples (1–5) with varying film compositions were used to obtain FeAlSi films with optimal compositions. The compositions of the as-deposited FeAlSi films were measured with inductively coupled plasma (ICP) spectrometry. Table I summarizes the composition (at. %) of each target and film, where the samples are arranged in order of the Al concentration and categorized into Al rich or Al poor. As shown, a composition deviation between the films and targets was observed, as reported previously.²¹ The thin-film composition of samples 1–3 differed despite having the same target composition; this was likely due to individual differences in the target fabrication process.

Figure 1(a) shows the XRD patterns (2θ–θ scan) of FeAlSi films annealed at T_a = 300–600 °C. Because there was no significant difference in the XRD results between the compositions, only the results for sample 3 are shown as a typical result. (The results for other composition samples are shown in the supplementary material Figs. S3 and S4.) FeAlSi(004) A2 peaks were observed for all annealed films, indicating the crystallization of FeAlSi. In addition, FeAlSi(002) B2 peaks were observed at annealing temperatures above 400 °C, indicating a change in the ordered structure above B2. Figure 1(b) shows the XRD patterns (φ scan) in which we observed FeAlSi(111) D0₃ peaks and verified a D0₃-ordered structure. Figure 1(c) shows the degree of order for D0₃ and B2 (S_D03, S_B2). S_D03 and S_B2 were calculated by using the intensity ratio between superlattice peaks (I₁₁₁-D0₃ and I₂₂₂-B2) and fundamental peaks (I₄₄₄-A2) observed in the (111)-plane ϕ scan XRD patterns, with the following formulas as a Ref. 22:

In Eq. (5), LP is the Lorentz-polarization factor of single crystals,²³ $ψ$ is the powder ring distribution factor,²⁴ and $Fhkl$ is the structure factor of the D0₃-FeAlSi unit cell for $(hkl)$ diffraction. The factor expressed as an exponential function is a Debye–Waller factor. The degree of D0₃ order increased monotonically as annealing temperature increased, while the degree of B2 order did not change much. The increase in the degree of D0₃ order is presumably due to the enhancement of atomic order by annealing. In other words, the overall trend is for the crystal structure to become more ordered as annealing temperature increases, approaching a perfect D0₃-ordered structure similar to that of bulk.

Figure 2(a) depicts the dependence of the magnetization curves measured by VSM on the annealing temperature, with the results of sample 3 shown as a typical example. (The results for other composition samples are shown in the supplementary material Fig. S5.) The two-step behavior of the magnetization curves for the as deposited samples is considered to be a phase separation, and the films are homogenized by annealing. Fluctuations in composition or degree of order may have caused phase separation and will need to be further investigated in the future. We added smaller scales of the magnetization curves to the inset only for those with small anisotropy. It is important to note that the magnetic easy axis changes as the annealing temperature increases. Specifically, below 400 °C, the [100] direction is the easy axis, while above 500 °C, the [110] direction is the easy axis. This indicates that the sign of the magnetocrystalline anisotropy constant K₁ has changed from positive to negative, and K₁ approaches zero at around 450 °C. The sign reversal of K₁ with annealing was also observed for samples 1 and 2 with Al-rich composition, which suggests that Al composition is important for attaining K₁ ∼ 0 in FeAlSi films. The annealing temperature at which K₁ approached zero varied slightly between 400 and 450 °C in the Al-rich samples, showing that the sign of K₁ and the temperature at which K₁ changes both depend on the film composition. Figure 2(b) shows the annealing temperature dependence of H_k for all samples with varying compositions estimated in magnetization curves, where we can see that H_k is minimal at around 400–450 °C for samples 1, 2, and 3. In contrast, samples 4 and 5 (films with poor Al) did not show K₁ ∼ 0. The lowest H_k value was below 1 Oe for sample 3, which is a comparable value to that of other soft magnetic materials such as NiFe or CoFeSiB.

These findings suggest that the K₁ ∼ 0 point differs between bulk and thin films, and the composition of K₁ ∼ 0 for films is Al-rich compared with that of bulk. This is presumably due to the volume balance of various ordered phases. As reported previously,²⁵ the soft magnetic property of bulk Sendust is composed of a complex structure of D0₃-Fe₃Si (K₁ > 0) and D0₃-Fe₃Al (K₁ < 0). When these structures are mixed in a Sendust central composition ratio, the K₁ of Sendust approaches zero. The degree of D0₃ order is almost 100% in the bulk, while in the film, it decreases, and disordered B2 and A2 structures are included, namely, D0₃-Fe₃Si (K₁ > 0), D0₃-Fe₃Al (K₁ < 0), B2-Fe₃Si (K₁ > 0), B2-Fe₃Al (K₁ > 0), and A2-FeAlSi (K₁ > 0).^25,26 Because D0₃-Fe₃Al only contains negative K₁, if the degree of order decreases from D0₃ to B2, a richer Al is needed to obtain K₁ ∼ 0. For a more quantitative discussion, we simulated the dependence of K₁ on the Al concentration and degree of order using the K₁ values reported previously.^25,26 Figure 3 summarizes the experimental and simulation results of the dependence of K₁ on the Al concentration for the compositions where K₁ ∼ 0 was attained. The solid lines show the simulated results for the dependence of K₁ on the Al concentration for the bulk and samples 1–3 films. The volume fraction of D0₃, B2, and A2 was assumed to be calculated by S_D03, S_B2-S_D03, and 100-S_B2, and the Fe composition was measured by the ICP method. The total K₁ was calculated as the sum of the products of the volume fraction and the reported value of K₁ for D0₃-Fe₃Si, D0₃-Fe₃Al, B2-Fe₃Si, B2-Fe₃Al, and A2-FeAlSi. The calculated results were roughly in agreement with the experimental results, and the points of K₁ ∼ 0 shifted to the Al-rich composition as atomic ordering decreased. The slight deviation between the calculation and experiment was caused by errors in the ordering parameter evaluation from the XRD results, the composition analysis by the ICP method, and the estimated volume fraction calculated by the ordering parameter. According to the previous studies on bulk and μm-order films with a perfect D0₃-ordered structure, the Sendust central composition needs to obtain an ideal soft magnetic property. Meanwhile, for nm-order films with a disordered structure, we found that the volume balance of D0₃-Fe₃Si, D0₃-Fe₃Al, B2-Fe₃Si, B2-Fe₃Al, and A2-FeAlSi was important for attaining K₁ ∼ 0. It is particularly important to control the volume fraction for D0₃-Fe₃Al, as it only yields negative K₁. In other words, in the case of nm-order FeAlSi films, K₁ ∼ 0 can be easily attained by controlling the film composition and atomic ordering while varying the sputtering target composition and annealing temperature. Our findings should prove useful for the application of FeAlSi films with optimal soft magnetic properties to highly sensitive TMR sensors.

We further clarified the magnetic properties of FeAlSi films prepared by sample 3 (T_a = 450 °C), which exhibited the most optimal soft magnetic properties in the VSM measurements. Figure 4 shows the angle dependence of the resonant field in FMR. We determined that the change in the resonant field was dominated by the fourfold component of the epitaxial thin film, and the contribution of the uniaxial component was very small. The fitted parameter values were K₁ = 1.6 × 10² (erg/cc), K_u = 10 (erg/cc), and $ω/γ$ = 88.5 (Oe). The observed K₁ value was comparable to that of bulk-FeAlSi in the previous study,²⁵ and we fabricated nm-order FeAlSi films with desirable soft magnetic properties. The exact value of H_k was quite small (0.43 Oe), indicating that the soft magnetic properties of the prepared FeAlSi film are more ideal than those of other typical soft magnetic materials such as NiFe and CoFeSiB.

Next, we discuss the uniaxial magnetic anisotropy constant K_u measured by FMR. The uniaxial strain in the in-plane direction of the film during the fabrication of the epitaxial film may have caused the uniaxial magnetic anisotropy as a result of the magnetostriction effect. To investigate this, we evaluated the magnetostriction constant and crystal strain by using the optical lever method and RSM, respectively, and the magnetic anisotropy constant due to magnetostriction was estimated using the following equation:

where $σ$ is the film strain.

Figure 5 shows the angle dependence of $δl/l110$ measured by the optical lever method. The results for sample 2 with T_a = 400 °C are shown as a typical example. The calculated λ₁₁₁ was 3.53 × 10⁻⁷, which is a similar value to that of bulk (10⁻⁷).²¹ The supplementary material Fig. S1 shows the results of in-plane RSM for sample 3 with T_a = 400 °C. The epitaxial growth of the FeAlSi film was observed, and the in-plane strain, which was calculated from the position of the (100) and (010) peaks, was σ = 2.06 × 10⁻⁵. The value of K_u calculated by Eq. (6) was 20.3 (erg/cc), which differs slightly from the value of the uniaxial component of the magnetic anisotropy constant observed by FMR but is almost identical. This suggests that the uniaxial in-plane magnetic anisotropy in FeAlSi films is due to magnetostriction, and that its contribution to the total magnetic anisotropy is small. Thus, we conclude that the fourfold symmetric magnetocrystalline anisotropy constant K₁ is dominant in the soft magnetic properties of FeAlSi films.

In summary, we fabricated epitaxial FeAlSi films on MgO substrates and verified that the annealed FeAlSi films contained a mixture of A2, B2, and D0₃ ordered structures. VSM measurements showed that the sign of the magnetocrystalline anisotropy constant K₁ changed as the annealing temperature increased in certain FeAlSi compositions. This is presumably due to the change in the volume balance of ordered phases caused by the annealing process and the point at which K₁ ∼ 0 shifts to the Al-rich concentration, decreasing the degree of D0₃ ordering. The exact value of K₁ measured by the FMR measurement was 159.3 (erg/cc), which is comparable to that of bulk, and the obtained H_k of 0.43 Oe was lower than that of other conventional soft magnetic materials. The uniaxial component of the magnetic anisotropy due to magnetostriction was negligible compared with magnetocrystalline anisotropy with fourfold symmetry. The developed FeAlSi films will be useful as a soft magnetic layer for highly sensitive TMR sensors. Although this study demonstrated the potential of FeAlSi epitaxial films for TMR sensors, further studies are needed to attain both high TMR ratio and soft magnetic properties in MTJ device structures.

See the supplementary material Fig. S1 for the RSM using an XRD pattern in-plane scan of (a) the entire reciprocal lattice space and (b) only the first quadrant, Fig. S2 for a schematic diagram of the FMR, and Figs. S3–S5 for the results of other composition samples (samples 1, 2, 4, and 5). In addition, we provided calculation details in a supplementary material to validate the calculations.

This work was partly carried out at the Fundamental Technology Center (Research Institute of Electrical Communication, Tohoku University) and supported by the Japan Society for the Promotion of Science (JSPS), the Japan Science and Technology Agency (JST) S-Innovation program, the Center for Science and Innovation in Spintronics (CSIS, Tohoku University), the Center for Spintronics Research Network (CSRN, Tohoku University), the Instrumental Analysis Group (Technical Division, School of Engineering, Tohoku University), the GP-Spin at Tohoku University, the ANRI Fellowship, and the Tohoku University (Professor Endo group).