Soft magnetic nanocomposite CoZrTaB–SiO₂ thin films for high-frequency applications

There has been a recent demand for miniaturization of transformers and inductors for power supplies, thanks to the ultra-high switching frequencies of power circuitry that require small energy storage and consequently, smaller size for the passive components.¹ The highest level of miniaturization will be achieved by the complete integration of the power supply onto silicon, i.e., Power Supply on Chip (PwrSoC); therefore, integration becomes more feasible.² This level of integration requires high-performance soft magnetic materials to operate at MHz switching frequency range and hence provide an opportunity for the integration of magnetic devices comprising magnetic films as the core.³ Current conventional power converters use relatively low flux density (≤0.5 T) ferrite magnetic cores, which makes them bulky and thus, offer a fundamental roadblock for the device miniaturization.¹ Consequently, there is a significant demand for soft magnetic materials exhibiting high saturation magnetization, low material loss, and medium to high permeability in MHz frequency range.⁴ 

Thanks to the unique disorder atomic structure of amorphous alloys, such as Co–Zr–Ta–B (CZTB),^4,5 which exhibit highly desirable soft magnetic properties, such as high saturation magnetization (M_s ∼ 1–1.5 T), ultra-low coercivity (H_c < 1 Oe), high resistivity (ρ > 115 μΩ cm), and tunable permeability (μ_r ∼ 300–900).^4,6 Most importantly, the films exhibit in-plane uniaxial magnetic anisotropy, ensuring that the magnetization process is dominated by coherent magnetization rotation, instead of domain wall displacement, to suppress the loss of the material at high frequencies.⁷ However, when the switching frequency of the devices approaches the MHz range, the eddy current losses, W_e, significantly dominate the hysteresis losses as W_e ∝ f², suggesting a dire need to develop novel materials with high resistivity to suppress the W_e.⁸ An extensive amount of work has been performed to understand the magnetic performance of Co–Zr–Ta–B alloy, mainly due to its interest at the industrial scale.^3,5 However, the resistivity of the alloy, which remained ρ ∼ 115 μΩ cm,⁹ did not emerge as a focus of interest. This study, particularly, focuses on improving the performance of material by tuning the electrical resistivity to address the eddy current loss at high frequencies. This might be possible by transforming the amorphous state of metallic Co–Zr–Ta–B films to the nanocomposite by introducing insulating phases. The addition of insulator is thought to form a dense amorphous atomic structure with three-dimensional insulating layers or boundaries leading to a higher resistivity.¹⁰ 

The motivation for the present study began with the need for developing a novel soft magnetic thin-film material aimed at significantly reducing eddy current loss by increasing its resistivity, which could make it a potential material for high-frequency thin-film inductors, writing heads, and magnetic sensing. Nanocomposite thin films of CZTB–SiO₂ were magnetron co-sputtered to investigate how SiO₂ content can affect the structural, electrical transport, dc magnetic, and finally yet importantly, high-frequency soft magnetic properties for MHz drive applications. The tuned inclusion of SiO₂ as an insulator, by varying the sputtering power, in the CZTB films was correlated to substantially improve resistivity and ultra-soft magnetic properties. The high-frequency permeability, ferromagnetic resonance (f_FMR), and damping constant (α) of CZTB–SiO₂ films are discussed in detail.

Amorphous thin films of CZTB–SiO₂ composite with varying concentrations of SiO₂ were deposited using DC and RF co-sputtering from Co₈₄Zr₄Ta₄B₈ (at. %) and SiO₂ targets, respectively, using the sputter deposition system (Lesker CMS-18). Prior to the deposition of films, the process chamber was pumped down to a base pressure of 10⁻⁷ Torr. Argon gas was introduced into the chamber to produce the desired sputtering pressure (i.e., 7 mTorr). The CZTB target was biased at 500 W with the SiO₂ target set at various powers from 100 to 250 W. Both targets (K. J. Lesker Company) were 3 in. in diameter, and the same targets were used for all deposition runs. The films were deposited onto 4-in. silicon wafers with a 0.25 μm silica layer. Prior to deposition, the substrates were pre-treated by using ion source radiation from a tungsten filament for 120 s An adhesion layer of 20 nm titanium was deposited prior to commencing the deposition of magnetic-dielectric composite films. Before the co-sputtering, the deposition rates of CZTB and SiO₂ were measured separately. The deposition rate of CZTB was confirmed as 20.8 nm/min by measuring the step height using a profilometer (KLA Tencor P17). Step heights of SiO₂ films were measured to calibrate the deposition rates at different biased powers (100 W and 250 W), as presented in Fig. 1. Uniaxial magnetic anisotropy was induced using alignment magnets during the deposition of films. The wafer was rotated (10 rpm) to achieve uniform film thickness, and the deposition was in a bottom-up configuration. Once deposited, the overall thickness of the films was measured.

The atomic structure of films was investigated using a x-ray diffraction technique (Philips Xpert diffractometer Cu − K_α λ = 1.54 Å). Scanning electron microscopy (SEM, Quanta 650) was used to image the surface morphology of films. The films were annealed at 550 °C in an Argon atmosphere with a heating rate of 10 °C/min for 1 h and subsequently characterized for their atomic structure. TEM lamella cross sections of the films were prepared using a dual-focused ion beam (FIB, FEI Helios NanoLab 600i). Microstructural analysis was performed using Selected Area Electron Diffraction (SAED) and a high-resolution transmission electron microscope imaging (HR-TEM Jeol 2100) with an electron beam of 200 kV.

The static magnetic properties of films were measured using a BH loop tracer (SHB Mesa-200) on 2 × 2 cm² diced samples. BH-loops were obtained when the films were saturated with a maximum field of 500 Oe. A vibrating samples magnetometer (VSM, Lakeshore) was used to investigate how magnetic saturation (M_s) was affected by the incorporation of the SiO₂ in CZTB films. The Curie temperature (T_c) of the material was investigated by magneto-thermo-gravimetry, which can sensitively pinpoint the T_c of amorphous and subsequent crystalline phases in the matrix.¹¹ The high-frequency permeability characteristics were determined on 4 × 4 mm² samples with a high-frequency permeameter (Ryowa-9M). Simulated material characteristics were obtained by solving the Landau–Lifschitz–Gilbert (LLG) equation, such as the dimensionless damping parameter (α), and fitted to the experimental data for comparison. The four-probe method was used to measure the resistivity (ρ) of the films with the contribution of the titanium adhesion layers’ resistivity being subtracted in each case.

Given that the films were in the range of ∼100–111 nm, the grazing-incidence small-angle XRD technique was utilized to investigate the atomic structure of the films. This involved positioning the beam source at 2.5° relative to the film surface. This reduces the penetration depth of x rays into the substrate and maximizes the Bragg reflections from the thin-film material. The resulting spectrum in Fig. 2 exhibited an overall amorphous nature for the as-deposited films. In the as-deposited state, the resulting patterns also show no presence of cobalt or cobalt oxide compounds which should be visible in the 2θ = 35–60° region if they were formed during co-deposition. Any crystalline phases formed would be detrimental, due to magnetocrystalline anisotropy and reduced resistivity, to its soft magnetic properties.

High-resolution TEM images of the cross section of metallic CZTB and nanocomposite CZTB–SiO₂ (250 W) films are presented in Figs. 3(a) and 3(b). A similar nanostructure was observed in both types of films, showing that the columnar growth mechanism remained dominant in the present series of films. Nevertheless, a slight increase in the width of the columns is obvious from the micrographs, suggesting the enhanced clustering of atoms by the co-sputtering of SiO₂. The same effect is more prominent from the high-resolution SEM image of the surface of films, as presented in Figs. 3(c) and 3(d). The columnar growth mechanism is well-known for high-pressure magnetron sputtered thin films.¹² The larger grains (∼40 nm) in CZTB–SiO₂ films, as compared to CZTB films (i.e., ∼31 nm), could be attributed to the many-element system occurring with denser atomic clusters, growing due to the addition of SiO₂ coupling to CZTB clusters.¹³ Furthermore, the SAED pattern of both types of films did not show any diffraction patterns; rather, it revealed diffused rings to confirm the amorphous nature of the films at the local scale. The columnar structure of the amorphous films is of significant importance as it provides an extra degree of electron scattering to enhance the electrical resistivity of the material.¹⁴ From the analysis of the growth rates, the at. % of SiO₂ in the material can be approximated, with a minimum of (CoZrTaB)₉₅–(SiO₂)₅ at 100 W to a maximum of (CoZrTaB)₈₉–(SiO₂)₁₁ at 250 W power.

DC magnetic properties of films were characterized to investigate if soft magnetic properties and, specifically, uniaxial anisotropy are well retained after SiO₂ inclusion in CZTB. Figure 4(a) presents a typical BH loop of the CZTB–SiO₂ (250 W) film along easy and hard anisotropy axis, showing well-defined magnetic anisotropy induced during the deposition process. Interestingly, the anisotropic field (H_k), which is a measure of uniaxial anisotropy, gradually increased from ∼32.5 Oe for pure CZTB to maximum ∼42.2 Oe for the CZTB–SiO₂ (250 W) film, as presented in Fig. 4(b). A well-defined uniaxial anisotropy is ideally required to retain the magnetization reversal by coherent rotation, instead of domain wall displacement, to suppress the materials’ loss for high-frequency drive applications. Furthermore, the coercivity (H_c) of the films along the hard anisotropy axis, which is a device operational direction for thin-film passives, gradually decreased from 0.6 to 0.1 Oe on SiO₂ inclusion, showing that nanocomposite films retained ultra-soft behavior even after introducing the highest fraction of SiO₂.

The co-sputtering of SiO₂ may introduce uniformly distributed non-magnetic insulating phase in a ferromagnetic amorphous matrix and according to the random anisotropy model (RAM), initially proposed by Zobin and Harris,¹⁵ for amorphous materials and later extended to two-phase nanocrystalline materials by Herzer et al.,¹⁶ the exchange coupling between the magnetic phases promotes the soft magnetic properties if the magnetic phases remain within the exchange-correlation length. While keeping magnetic phases in exchange-correlation length (i.e., 5–10 nm for Co), the H_c of films further improved due to the homogeneous distribution of non-magnetic SiO₂ at higher sputtering power.¹⁷ At the same time, the incremental H_k as a function of SiO₂ could be attributed to the weak magnetic coupling between the ferromagnetic phases attained on the larger separation between magnetic phases. Jiang et al. reported the similar behavior of H_c for the addition of N₂ to the FeCo thin-film system, whereby the individual magnet moments become less magnetically coupled when the insulating phase becomes larger.¹⁸ Similarly, increasing the non-magnetic phase considerably increases the H_k of a magnetic composite material due to weak exchange coupling,^19,20 specifically, where there is an increased density of an amorphous metal oxide.²¹ 

The inset of Fig. 4(a) shows how M_s of CZTB–SiO₂ films gradually decreased by SiO₂ incorporation in CZTB films. The incremental SiO₂ in CZTB lowers the corresponding M_s. This is due to the dilution of magnetic elements in the thin-film system. This trend is in good agreement with the work of Munakata et al., where an increasing SiO₂ (vol. %) reduced M_s by a factor of two.²² In the present case, the materials’ ferromagnetic content density decreased on SiO₂, hence individually coupled magnet moments are decreased in value (assuming a homogeneous distribution). As the material becomes a multi-element system, it follows the standard ferromagnetic models; one might situate it gradually decreasing on the Slater–Pauling curve and its magnetic moments being dependent on Co content.

The electrical resistivity (ρ) of films increased by incorporating SiO₂ in CZTB films, as presented in Fig 5. The ρ of the metallic CZTB films was measured as ∼160 μΩ cm while it exponentially increased to maximum ∼224 μΩ cm for nanocomposite CZTB–SiO₂ films deposited with 250 W power. Interestingly, the ρ of CZTB films is significantly higher than the reported values (i.e., ∼115 μΩ cm) in the literature.⁹ The higher value of ρ, in the present case, could be attributed to the extra degree of electron scattering at the grain boundaries of columnar nanostructure (see Fig. 2), as discussed in Sec. II.²³ 

Depositing a dielectric in a system such that it is concentrated between short-range magnetic phases can reduce overall conductive properties and increase the resistivity.¹⁹ For example, in high oxygen content CoFeB–O films, the surrounding non-magnetic at. % follows an apparent proportional effect with resistivity as the weakly metallic oxide insulating layers increase in size, and metallic conductance effects become minimal.²⁴ A similar trend has been observed with the deposition of a nitride; thereby increasing its concentration increases the size of the amorphous space between magnetic particles and improving overall resistivity.¹⁸ Below a critical concentration of the conductor, larger resistivity is related to higher non-metallic phases in the metal–insulator–metal model assuming a heterogeneous distribution of magnetic and non-magnetic phases.¹⁰ Above a critical concentration, resistivity has been reported to increase exponentially as the conductivity begins to show tunneling effects.^10,25 Conclusively, the substantial increase in ρ, in the present case, corresponds to increasing SiO₂ content in the nanocomposite CZTB–SiO₂ films.²¹ 

The temperature dependence of magnetization, M(T), of CZTB–SiO₂ thin films was investigated using the magneto-thermo-gravimetric technique²⁶ (MTG), as presented in Fig. 6. This involves placing a sample with an arbitrary magnetic weight (M_w), inside a furnace with an applied gradient magnetic field. The value of its force is recorded and plotted as a function of temperature. As shown in Fig. 6(a), the material follows a standard ferromagnetic decay until it reaches a point indicated by T_C1, upon which it encounters a steady magnetization response for some temperature interval and then finally approaches zero at T_C2. The Curie points T_C1 and T_C2 were derived at the steepest temperature derivative, dM/dT, and presented in Fig. 6(b), which show a continuous decrease in the Curie points (T_C1 = 450–275 °C and T_C2 = 640–620 °C) as a function of SiO₂.

The two-phase M(T) behavior indicates that either film has a heterogeneous amorphous structure with two Curie points (T_C1, T_C2) or the Curie point of the amorphous phase is less than the crystallization temperature (T_x) of the films where new phases precipitate with a higher Curie point (T_C2 > T_C1).²⁷ This can be resolved by annealing the films in the temperature interval T_C1 − T_C2. Therefore, films were annealed at temperature T_a = 550 °C (i.e., T_C1 < T_a < T_C2) for 60 min and characterized for their atomic structure, as shown in Fig. 6(c). The XRD and SAED pattern, see Fig. 7, confirmed the crystallization of the amorphous films at 550 °C. Conclusively, the films were monolithic amorphous in the as-deposited and T_x of the CZTB–SiO₂ films is below the Curie point of the amorphous phase (i.e., T_C1). The new crystallized structure may have a higher Curie constant as discussed earlier, depending on its intrinsic properties that govern its decay, possibly explaining the plateau of the Curie curve. It then finally becomes non-magnetic between 600 and 650 °C. Approaching the crystallization temperature of CZTB/CZTB–SiO₂, different crystalline or nanocrystalline phases may precipitate and show entirely different global magnetic behaviors due to intrinsic properties, such as a magnetocrystalline anisotropy, higher dipole moment density, and a stronger exchange interaction.²⁸ 

Furthermore, there is an evident trend in a reduction of Curie points (both T_C1 and T_C2) with increasing SiO₂ power. The Curie point of the magnetic materials depends on the strength of exchange coupling between the magnetic entities. The magnetic exchange coupling gets weaker as the content of SiO₂ increases in the films, and consequently, it decreases the T_C1 of the amorphous phase. This finding is in good agreement with the incremental H_k values, which gradually increased on SiO₂ incorporation due to the weaker exchange coupling [see Fig. 4(b)]. More interestingly, the T_C1 approaches as low as ∼250 ^oC, suggesting that the small incorporation of SiO₂ makes a significant change in the magnetic ordering temperature. Furthermore, the gradual decrease in area under the first kink of M(T) curve indicates how volume fraction of the insulating phase in the CZTB films increased as a function of SiO₂ power. Moreover, the decrease in T_C2 as a function of SiO₂ is minimal, showing the precipitation of the same crystalline phases during heating for all films.

The high-frequency permeability spectrum of the films is presented in Fig. 8. The μ_r of the films gradually decreased as a function of SiO₂. With the described sputtering parameters, the as-deposited pure CZTB films attained a permeability of μ_r ∼ 328 at an arbitrary ∼100 MHz working frequency. Thereafter, with increasing SiO₂ power, the permeability of CZTB–SiO₂ dropped to a minimum μ_r ∼ 250 for 250 W SiO₂ power. The μ_r of CZTB–SiO₂ films was expected to decrease on SiO₂ incorporation due to the gradual increase in H_k [see Fig. 4(b)]. The SiO₂ also effected the full-width half maximum (FWHM) of ferromagnetic resonance peaks, which gradually reduced. A small shift toward lower frequencies in the ferromagnetic resonance frequency (f_FMR) of the films was found on the SiO₂ incorporation possibly due to the slight decrease in the materials M_s as seen in Kittel's equation.²⁹ The gradual decrease of M_s may also be responsible for the corresponding lower permeability values as, in general, M_s ∝ μ_r, being dependent on a materials’ ferromagnetic content. In addition, the imaginary component of permeability (μ″) represents the loss of the material, remained constant until reaching the f_FMR. This exciting behavior could be attributed to the small eddy current and anomalous loss, which are expected to be minimal due to significantly high resistivity (see Fig. 5) and magnetization reversal by rotation, due to well-aligned uniaxial anisotropy, along the hard magnetic anisotropy axis.

Further analysis involved fitting a permeability model obtained from solving the Landau–Lifschitz–Gilbert equation to the experimental data and extracting parameters such as the damping factor, α, which hinders magnetization reversal to explain the diminishing FMR peaks. The theoretical model is a function of several known properties such as M_s, H_k, α, and γ, the gyromagnetic ratio, making it trivial to isolate α.³⁰ The α of pure CZTB was calculated to a minimum of ∼0.14, which remained almost the same for the CZTB–SiO₂ at 100 W sputter power. However, an abrupt increase in α was found for the high content of SiO₂ films with a maximum ∼0.22 for the 250 W SiO₂ film. The damping loss, which is directly related to α, is considered the major source of material loss at high frequencies after eddy and anomalous loss.

It is evident that the addition of SiO₂ increases the dimensionless damping parameter and may subsequently affect losses during magnetization reversal in accordance with the LLG model. The contribution to the effective damping parameter usually arises from intrinsic and extrinsic conditions. In the present amorphous case, there is a minimal intrinsic effect, as there is no spin–orbit coupling to lattices in amorphous thin films. However, in the extrinsic case, magnon scattering and anisotropy dispersion are strongly related to α_. The line-broadening of the FWHM in the μ_r″ curves and its deviation from a standard Lorentzian suggest that anisotropy dispersion may be responsible for the increasing α. Conclusively, the inclusion of SiO₂ might produce small anisotropy dispersion from 90° hard–easy axis alignment and works as a fundamental source for increasing α. This has been seen with similar SiO₂ studies with the FWHM increasing by a factor of two.³¹ Comparable properties have been seen before with oxide rich phases such as insulating HfO₂ rich layers,¹⁹ in which the permeability dropped to μ_r = 75. These rich oxide phase materials may have their effective high-frequency properties due to their high resistivity and increased insulator phase separation between high moment CoFe(110) phases. In the present case, the incorporation of SiO₂ transforms the films from an amorphous metallic nature to the nanocomposite state. This might cause fluctuations in the internal magnetic field potential encountered by magnetization rotation during a reversal, which might be a result of an uneven molecular distribution of SiO₂ throughout the CZTB matrix. These AC results show how CZTB–SiO₂ may be applicable in low footprint inductors as it holds a moderate permeability and high M_s at high frequencies (up to 800 MHz) compared to conventional materials such as ferrites which show a permeability drop off at approximately 1–5 MHz (Fig. 9).

Nanocomposite thin films of CoZrTaB–SiO₂ were investigated for their high-frequency soft magnetic performance for passive magnetic components. This was achieved by co-sputtering films onto silicon wafers at various SiO₂ powers, characterizing their structure and thin-film growth mechanism, investigating uniaxial anisotropy and DC soft magnetic performance, exploring thermomagnetic behaviors, probing electrical resistivity and finally, AC soft magnetic performance up to GHz frequency range. The nanocomposite films remain amorphous while retaining well-aligned uniaxial magnetic anisotropy. The magnetic softness along hard anisotropy axis improved to H_c ∼ 0.1 Oe and anisotropic field significantly increased to maximum H_k ∼ 42 Oe on SiO₂ incorporation were attributed to the weaker exchange coupling attained on SiO₂ incorporation. The higher resistivity, ρ = 160 μΩ cm, of the Co–Zr–Ta–B films, as compared to the literature values (i.e., 115 μΩ cm), was attributed to the high degree of electron scattering at the boundaries of the clusters. Most importantly, the resistivity of CZTB–SiO₂ further improved to ∼224 μΩ cm with increasing SiO₂ content, attributed to the metal–insulator–metal network-like structure. Finally, the ac soft magnetic properties of the films show tunable permeability (μ_r ∼ 250–328) along with flat frequency response, deficient magnetic loss, and ferromagnetic resonance frequency (f_FMR > 1.2 GHz). The damping factor, α, of films was calculated solving LLG equation, which increased with the SiO₂ incorporation, and could be attributed to the extrinsic effects. The proposed approach of increasing the resistivity, ρ, of CoZrTaB films makes the material useful for high-frequency drive applications.

The authors acknowledge Science Foundation Ireland for funding this project (No. R17121). The author would like to thank Brendan Sheehan and Michael Schmidt in the Speciality Products and Services team in Tyndall, who provided TEM and SAED imaging for this publication. Special thanks to E. Dastanpour and V. Ström at KTH-Royal Institute of Technology, Stockholm, Sweden for the MTG measurements of the samples.

The data that support the findings of this study are available from the corresponding author upon reasonable request.