Mn-Zn ferrite samples prepared by conventional solid state reaction method and sintering at 1325 °C were Co-enriched by addition of CoO up to 6000 ppm and characterized versus frequency (DC – 1GHz), peak polarization (2 mT – 200 mT), and temperature (23 °C – 120 °C). The magnetic losses at room temperature are observed to pass through a deep minimum value around 4000 ppm CoO at all polarizations values. This trend is smoothed out either by approaching the MHz range or by increasing the temperature. Conversely, the initial permeability attains its maximum value around the same CoO content, while showing moderate monotonical decrease with increasing CoO at the typical working temperatures of 80 – 100 °C. The energy losses, measured by a combination of fluxmetric and transmission line methods, are affected by the eddy currents, on the conventional 5 mm thick ring samples, only beyond a few MHz. Their assessment relies on the separation of rotational and domain wall processes, which can be done by analysis of the complex permeability and its frequency behavior. This permits one, in particular, to calculate the magnetic anisotropy and its dependence on CoO content and temperature and bring to light its decomposition into the host lattice and Co2+ temperature dependent contributions. The temperature and doping dependence of initial permeability and magnetic losses can in this way be qualitatively justified, without invoking the passage through zero value of the effective anisotropy constant upon doping.
I. INTRODUCTION
The introduction of additional oxides is the classical strategy adopted for improving the broadband soft magnetic properties of Mn-Zn ferrites. One aims in this way at reduced eddy current losses, increased frequency stability of the permeability and optimal response at the typical working temperatures of the device cores.1,2 The main physical mechanisms by which the dopants can affect the ferrite magnetic properties consist either in the segregation of the foreign species at the grain boundaries or their dissolution inside the grains, thereby modifying the resistivity, the intrinsic magnetic properties, and possibly the microstructure of the material.3–5 It is thus standard practice, for example, to introduce CaO and Nb2O5 in order to modify the chemistry at the grain boundaries and increase the related resistivity, an effect that can be reinforced by adding SiO2 or Ta2O5.6,7 At the same time, with additives like TiO2, SnO2, and CoO and the corresponding substitution of Fe3+ cations with the foreign cations, the mechanism of hopping conduction between Fe2+- Fe3+ ions can be hindered and the intragrain resistivity can increase.3,8–10 In addition, the phenomenon of anisotropy compensation, where the negative anisotropy of the Fe3+ ions in the octahedral sites is made to combine with the positive anisotropy of the extra-cations occupying the same sites, is expected to occur, thereby lowering, in a temperature-dependent fashion, the overall magnetocrystalline anisotropy.11 Standard Mn-Zn ferrites generally rely on anisotropy compensation by extra Fe2+ ions, an effect that can be modulated by the additives, eventually affecting to large extent energy loss and permeability at all temperatures. These concepts are relatively well understood at a qualitative level, but hard to translate into a quantitative assessment, because this would require precise description of the magnetization process and the mechanisms of magnetic energy losses. The very low values of the anisotropy constant attained in these materials can indeed strongly enhance the rotational processes, if compared with the conventional steel sheets. On the other hand, the semi-insulating character of these materials naturally leads, depending on sample size and magnetizing frequency, to consider the dissipation channel associated with spin damping.
In this work we have investigated the effect of CoO doping on the properties of Mn-Zn sintered ferrites and developed a method to recognize the ensuing evolution of the permeability and energy loss behavior in terms of anisotropy evolution. This method takes at face value the broadband frequency response of the material by modeling the frequency dependence of permeability and losses and separating the contributions to the magnetization process of domain wall (dw) displacements and rotations. Rotation modeling is based, in turn, on the solution of Landau-Lifshitz equation and averaging under assumed distribution of the local (at the scale of the single grain) anisotropies. With the DC rotational susceptibility connected, for a uniform orientation distribution of the easy axes, to the average anisotropy K as χDC,rot = Js2/3μ0|K|, we eventually obtain, knowing the dependence of the saturation polarization Js on CoO content cCo (0 ≤ cCo ≤ 6000 ppm) and temperature T, the dependence of the effective anisotropy constant on cCo and T. By denoting K as effective, we take into account that the torque involved with rotational processes involves both magnetocrystalline and local magnetostatic effects. It is thus verified that the calculated K(cCo) is minimum for cCo = 3000 – 4000 ppm, corresponding to minimum magnetic losses and maximum initial permeability. K(cCo, T) shows a monotonical decrease with increasing temperature. The secondary maximum of the initial permeability, observed around T = 90 °C, turns then out to be related, according to the previous expression for χDC,rot, to the competition between the decreasing behaviors of Js and K with the temperature. In any case, the estimated anisotropy contribution by the Co2+ cations KCo(T) = K(cCo, T) - K(cCo= 0, T) appears to engender a decrease of K(T) below about 60 °C only for cCo < 5000 ppm.
II. SAMPLE PREPARATION. STRUCTURAL, ELECTRICAL, AND MAGNETIC CHARACTERIZATION
Mn-Zn ferrite powders were prepared following the conventional method of solid state reaction, starting from high-purity raw materials. Dry mixing of the Fe, Mn, and Zn oxides was performed using a laboratory mixer and prefiring took place at 850 °C in air. The resulting powders had the proportions: Fe2O3 = 72 wt%, MnO = 21.60 wt%, ZnO = 6.35wt %. The prefired powders were doped with a fixed doping scheme that included CaO and Nb2O5, while Co doping varied between 0-6000 ppm of CoO, step 1000. After milling, roll granulation was performed using a rotating drum with the addition of 0.2 wt% of aqueous solution of polyvinyl alcohol (Merck, Analytical Grade, MW72000). Specimens of toroidal shape of outside diameter of 14 mm were uniaxially pressed to a press density of 2.90 g/cm3 and sintered at a top temperature of 1325 °C under controlled partial oxygen pressure, according to the Morineau-Paulus equilibrium condition. The sinter density of the Mn-Zn ferrites was measured by the Archimedes’ method. The polycrystalline microstructure of the sintered specimens was investigated By Scanning Electron Microscopy (JEOL 6300), after grinding, polishing and chemical etching of the sintered ferrites. The X-Ray Diffraction analysis showed that in all the prefired powders the initial spinel structure was successfully formed. As shown in Table I, the sinter density of the compacted and sintered toroidal specimens, measured by Archimedes’s method, was observed to increase slightly with CoO doping, up to 2.4 % with CoO = 6000 ppm. The micrographs obtained by SEM-Jeol 6300 images of the polished and etched sample surface showed little variation of the average grain size, which attained a maximum value of 13.1 μm for CoO = 4000 ppm (see Table I and Fig. 1).
. | . | . | . | . | Real resistivity . | Imaginary resistivity . |
---|---|---|---|---|---|---|
. | CoO . | Grain size . | Density δ . | DC resistivity . | at 10 MHz . | at 10 MHz . |
Sample . | (ppm) . | <s> (μm) . | (kg/m3) . | ρDC (Ω m) . | ρ’10MHz (Ω m) . | ρ”10MHz (Ω m) . |
A0 | 0 | 10.8 | 4.92 | 4.42 | 0.0130 | -0.0076 |
A1 | 1000 | 10.7 | 4.98 | 5.08 | 0.0170 | -0.0080 |
A2 | 2000 | 11.9 | 4.98 | 5.39 | 0.0165 | -0.0073 |
A3 | 3000 | 12.2 | 5.01 | 5.52 | 0.0195 | -0.0069 |
A4 | 4000 | 13.1 | 4.98 | 5.66 | 0.0185 | -0.0069 |
A5 | 5000 | 10.8 | 5.01 | 5.9 | 0.0154 | -0.0076 |
A6 | 6000 | 11.1 | 5.04 | 6.1 | 0.0137 | -0.0096 |
. | . | . | . | . | Real resistivity . | Imaginary resistivity . |
---|---|---|---|---|---|---|
. | CoO . | Grain size . | Density δ . | DC resistivity . | at 10 MHz . | at 10 MHz . |
Sample . | (ppm) . | <s> (μm) . | (kg/m3) . | ρDC (Ω m) . | ρ’10MHz (Ω m) . | ρ”10MHz (Ω m) . |
A0 | 0 | 10.8 | 4.92 | 4.42 | 0.0130 | -0.0076 |
A1 | 1000 | 10.7 | 4.98 | 5.08 | 0.0170 | -0.0080 |
A2 | 2000 | 11.9 | 4.98 | 5.39 | 0.0165 | -0.0073 |
A3 | 3000 | 12.2 | 5.01 | 5.52 | 0.0195 | -0.0069 |
A4 | 4000 | 13.1 | 4.98 | 5.66 | 0.0185 | -0.0069 |
A5 | 5000 | 10.8 | 5.01 | 5.9 | 0.0154 | -0.0076 |
A6 | 6000 | 11.1 | 5.04 | 6.1 | 0.0137 | -0.0096 |
The electrical impedance was measured from DC to 20 MHz by means of a standard four-wire method, with good voltmetric and amperometric electrical contacts ensured by indium strips sandwiched, at the ends of a cut ring sample, between the polished sample surface and copper leads subjected to uniform pressure. The measuring method and a general discussion of the frequency dependence of the complex resistivity ρ(f) = ρ’(f) + jρ”(f) in typical ferrites are illustrated in Ref. 12. It is found in the present case that the real part ρ’ of the resistivity exhibits the usual double-plateau dependence on frequency,12 with the DC value ρDC monotonically increasing from 4.42 Ωm to 6.1 Ωm on passing from the undoped to the 6000 ppm CoO doped ferrite and the high-frequency plateau (observed beyond about f = 1 MHz) attaining a value around 0.015 – 0.020 Ωm. The imaginary part ρ”displays the well known capacitive character, as dictated by the thin insulating grain boundary.12 The DC resistivity ρDC is thus overwhelmingly associated with the grain boundaries, which are capacitively short-circuited at high frequencies, so that the high-frequency resistivity basically coincides with the grain resistivity. The evolution of DC resistivity and the real and imaginary resistivities at 10 MHz versus CoO doping is shown in Table I.
The magnetic characterization of the sintered ring samples (outside diameter 14.15-14.44 mm, inside diameter 9.15 – 8.86 mm, thickness 4.55 – 5.02 mm) was performed by means of a calibrated hysteresigraph-wattmeter from DC to a maximum frequency of 10 MHz and from a few hundred kHz to 1 GHz by a transmission line method using a Vector Network Analyzer (VNA, Agilent 8753E, 10 mW irradiated power) and a 50 Ω coaxial line. The ring sample was placed against the shorted bottom of a clean 50 Ω brass cell and the complex permeability was obtained from the measurement of the scattering parameter in reflection. The fluxmetric and transmission line methods provided overlapping results in the common frequency interval. The measuring method, setup, and procedure are discussed in detail in Refs. 13,14. Measurements as function of temperature could also be performed up to 120 °C, as discussed in Ref. 15. The saturation polarization Js as a function of temperature was in particular obtained. Table II summarizes the behavior of significant magnetic parameters as a function of CoO content. It is noted in Tables I and II and shown in Fig. 2 that the saturation polarization Js and the grain resistivity (i.e. ρ’ at 10 MHz) concurrently pass through maximum and minimum values value for a CoO content cCo = 3000 – 4000 ppm. The increase of the grain resistivity with the introduction of the CoO is consistent with the idea that the Co2+ cations entering the host lattice prevalently occupy the octahedral sites and take the place of either Fe2+ or Fe3+ ions (which are endowed with higher magnetic moment), thereby hindering electron hopping between Fe2+ and Fe3+ ions.8,9 It appears, however, that this actually would occur only for cCo < 4000 ppm, because ρ’is observed to decrease again at high CoO contents, in association with an increase of Js. The DC resistivity is instead observed to always increase with cCo, pointing to correspondingly increasing dopant segregation at the grain boundaries. There is no unambiguous explanation for the concurrent behaviors of grain resistivity and saturation polarization at the highest cCo values. These could be possibly due to some kind of sharing of the Fe and Co cations between octahedral and tetrahedral sites. In any case, by adding dopants, like CoO, soluble in the spinel lattice, anisotropy compensation expectedly occurs. Indirect evidence for it is given by the evolution of initial permeability and magnetic losses with dopant content and temperature T.8–11 This effect is well known and discussed in the literature, but no direct determination of the anisotropy evolution is provided.16 We shall see in the following that, starting from analysis and modeling of the frequency dependence of permeability and losses, we are eventually able to describe the evolution of the anisotropy constant K with cCo and T, and to single out the contributions from the host lattice and from the doping ions.
Sample . | CoO (ppm) . | Js (T) . | μr,2mT . | μ’r,100mT . | W100mT(J/m3) . | μDC,rot . | K (J/m3) . |
---|---|---|---|---|---|---|---|
A0 | 0 | 0.540 | 2230 | 4199 | 1.540 | 2110 | 36.68 |
A1 | 1000 | 0.534 | 2258 | 3889 | 1.310 | 2070 | 36.56 |
A2 | 2000 | 0.529 | 2480 | 3906 | 1.020 | 2210 | 33.61 |
A3 | 3000 | 0.524 | 2690 | 3742 | 0.845 | 2380 | 30.38 |
A4 | 4000 | 0.522 | 2555 | 3401 | 0.687 | 2300 | 31.44 |
A5 | 5000 | 0.532 | 2070 | 3712 | 1.185 | 1713 | 43.85 |
A6 | 6000 | 0.536 | 1385 | 3100 | 2.100 | 1190 | 64.07 |
Sample . | CoO (ppm) . | Js (T) . | μr,2mT . | μ’r,100mT . | W100mT(J/m3) . | μDC,rot . | K (J/m3) . |
---|---|---|---|---|---|---|---|
A0 | 0 | 0.540 | 2230 | 4199 | 1.540 | 2110 | 36.68 |
A1 | 1000 | 0.534 | 2258 | 3889 | 1.310 | 2070 | 36.56 |
A2 | 2000 | 0.529 | 2480 | 3906 | 1.020 | 2210 | 33.61 |
A3 | 3000 | 0.524 | 2690 | 3742 | 0.845 | 2380 | 30.38 |
A4 | 4000 | 0.522 | 2555 | 3401 | 0.687 | 2300 | 31.44 |
A5 | 5000 | 0.532 | 2070 | 3712 | 1.185 | 1713 | 43.85 |
A6 | 6000 | 0.536 | 1385 | 3100 | 2.100 | 1190 | 64.07 |
III. PERMEABILITY, MAGNETIC LOSSES, AND THEIR ANALYSIS. ROLE OF DOPANT CONTENTS AND TEMPERATURE
A main aspect of the Mn-Zn ferrite phenomenology is represented by the loss and permeability response of the material as a function of the magnetizing frequency. One chief objective of ferrite development is then one of broadening, within the constraints given by the Snoek’s law, the useful frequency band and of minimizing the energy losses at the typical working temperatures.7 The strategy based on doping aims, besides the increase of the grain boundary resistivity, at modulating the magnetocrystalline anisotropy and its dependence on temperature, with ensuing effects on loss and permeability. The mechanism of anisotropy compensation is, in particular, expected to operate with Co2+ ions, but no quantitative information on the anisotropy behavior is so far available.
A. Magnetic permeability and losses. Rotations versus domain wall displacements
Our experimental approach relies first on a complete determination of complex permeability μ = μ’ - jμ” and energy loss W, up to the frequencies at which both μ’ and μ” become vanishingly small. Fig. 3 provides an example (A2 sample, CoO = 2000 ppm) of real permeability behavior measured up to 1 GHz for peak polarization value Jp ranging between 2 mT and 200 mT. As discussed in previous papers12,17,18 and shown in this figure, the VNA measurements, though performed under defined power (10 mW), overlap with the fluxmetric measurements made at different Jp values at sufficiently high frequencies. This fact is explained in terms of full relaxation of dw processes and uniquely surviving rotational processes at such frequencies.18 It is also noted that μ’ is basically constant up to about 100 kHz, with resonant hump and relaxation decay occurring in the MHz region. The energy losses at given Jp value are fluxmetrically measured up to 10 MHz. They are instead calculated, when adopting the VNA method, from the measured complex permeability, according to
Fig. 4 shows overlapping of the fluxmetric (symbols) and VNA (continuous lines) energy loss results, paralleling the permeability overlap observed in Fig. 3. On increasing the Jp value, the width of the overlapping frequency interval decreases, because the dw displacements increase their contribution, at the expenses of rotations. Practical measuring limitations do not permit one to fluxmetrically attain the MHz range beyond some 50 mT, but reasonable extrapolation to the high frequency VNA loss values can be made (short-dash lines in Fig. 4). By the VNA measurements, where the exciting field amplitude decreases with frequency, we almost exclusively excite the rotational processes. Branching of the VNA and fluxmetric loss curves (shown, for example, in Fig. 4 for Jp = 50, 100, and 200 mT between 100 kHz and 1 MHz) provides experimental evidence of the respective role played by the dw and the rotational losses. The shown W(f) curves refer to the as-obtained materials. By repeating the measurements after grinding the ring samples down to about 1 mm thickness, we conclude, according to the example shown for Jp = 10 mT in Fig. 4, that the eddy currents can contribute to the energy loss, by an amount of the order of 10 - 20 %, only beyond a few MHz, in close analogy with previous experiments performed on Mn-Zn ferrites of different tickness.18
It has been shown in previous papers that the whole phenomenology of permeability and losses in Mn-Zn ferrites can be assessed to very good extent by detailed analysis and modeling of the magnetization process and the related dissipation mechanisms.12,17,18 A fundamental approximation of modeling consists in separating the contributions of dw displacements and rotations, generalizing to poorly conducting materials the basic assumptions of the statistical theory of losses.19 The rotations now parallel the mechanisms underlying the classical losses in magnetic sheets, while the local character of the dissipation by the dws, whose dynamics can be described by the very same equations in conductive and in insulating magnetic materials,20 is at the root of the hysteresis and excess losses. A procedure for enucleating dw and rotational permeabilities from the measured permeability versus frequency behavior can actually be worked out by a self-consistent procedure. This important aspect of the magnetization process is discussed in detail in Ref. 18. The method exploits the fact that the imaginary permeability at low-frequencies is exclusively associated with the dw processes μ”DC μ”DC,dw. On the other hand, the rotational permeability can be considered, for an assumed isotropic distribution of easy axes, quite independent of Jp, at least up to Jp ∼ 0.1 Js – 0.2Js. It is thus demonstrated in Ref. 18 that these two assumptions suffice to separately calculate μ”dw(Jp, f), μ’dw(Jp, f), μ’rot(f), and μ”rot(f). Figs. 5a and 5b provide an example, concerning the A4 sample (CoO = 4000 ppm), of permeability decomposition versus frequency. It is observed here that the initial permeability (the one measured, in our case, with Jp = 2 mT) is overwhelmingly due to the rotational processes. The Jp-independent μ’rot(f), and μ”rot(f) curves exhibit the expected resonant behavior, in contrast with the relaxation character of μ’dw(Jp, f) and μ”dw(Jp, f). The overall dependence of the measured DC-permeability μDC on the CoO content in the Jp interval 2 mT - 200 mT is shown in Fig. 6. The singled out rotational contribution μDC,rot is observed to pass through a maximum value for CoO = 3000 – 4000 ppm. A more or less monotonical decrease of μDC with CoO content is observed beyond about Jp = 50 mT, where the dw contribution becomes prevalent. To note the small difference, consistent with the results shown in Fig. 5a, between μDC and μ’(100 kHz). Fig. 6b shows that a greater stability of the initial permeability versus temperature is obtained for CoO = 3000 - 4000 ppm.
The dependence of permeability on cCo shown in Fig. 6 is paralleled by the behavior of the energy loss W(f, Jp) shown in Fig. 7. This example refers to Jp = 100 mT and the frequency range 100 Hz - 800 kHz. The softening role of CoO addition is apparent up to about 4000 ppm, where W(f) and μDC,rot (Fig. 6) start to increase and decrease, respectively. Weakening dependence of W(f) on CoO is observed on approaching the MHz range. The dw processes tend to earlier relaxation compared to rotations (see Fig. 5) and the rotational losses become prevalent. The quantitative theoretical treatment of the latter, based on the application of Eq. (1), using the real and imaginary rotational permeability components as solution of the Landau-Lifshitz equation,12,17,18 actually predicts that all loss versus frequency curves eventually coalesce, for a given Jp value, on going deep into the MHz range.
B. Effect of doping on the anisotropy. Temperature dependence of loss and permeability
Experimental knowledge of the rotational permeability permits one to obtain good insight on behavior and role of the magnetic anisotropy, whose manipulation is one key objective of doping. If we assume, in fact, that in the polycrystalline specimen the easy axes are uniformly distributed in the half-space and if we denote with K the average anisotropy constant, including the effects of the local demagnetizing fields, we estimate, with reasonable approximation, the DC relative rotational permeability of the ferrite sample as
and we obtain the K (conventionally taken as positive) versus the CoO content shown in Fig. 8. It is generally assumed that by doping one might ideally achieve near-zero anisotropy, the result of compensation of the negative anisotropy of the host by the positive anisotropy of the added cations. The temperature at which full compensation expectedly occurs, with ideal passage of K through zero value and ensuing benefits in terms of increased permeability and lowered losses, is a function of the dopant concentration, which can be adjusted in order to achieve the best material response at a convenient temperature. Regarding the specific case of CoO doping, the temperature T at which minimum losses and maximum initial permeability are measured is generally found to decrease with increasing CoO content.9,10 An example of energy loss dependence on temperature in the present materials is shown in Fig. 9, concerning a typical working condition, where Jp = 100 mT and f = 100 kHz. It is found that W(f) attains a minimum value between 80 °C and 100 °C, but for cCo = 4000 ppm. For this doping amount, providing minimum loss at room temperature (Fig. 9b), we obtain the greatest thermal stability of W(f) beyond room temperature. Anisotropy evolution with CoO and temperature is necessarily involved in such a behaviour, in combination with structural effects (pinning centers, grain boundaries, porosity). This is qualitatively recognized in the literature, but quantitative information is lacking. By singling out the rotational permeability μDC,rot, as shown in Fig. 5, and by determining the saturation polarization Js as a function of temperature, as shown in Fig. 10, we can estimate, by means of Eq. (2), the doping-dependent magnetic anisotropy K(T). The temperature dependence of μDC,rot is shown in Fig. 11. It is noted that the secondary maximum of μDC,rot versus temperature, usually assumed as due to the passage of K(T) through zero value, can be better understood, according to Eq. (2), in terms of competing decreasing behaviors of Js and K with the temperature. We find in fact that K(T) is a CoO dependent monotonically decreasing function of temperature, as shown in Fig. 12. By taking then the difference between the calculated anisotropy constants for doped and undoped ferrites, we obtain the behavior of the additional term introduced by the Co2+ ions KCo(T) = K(cCo, T) - K(cCo= 0, T). This is shown in Fig. 13. One can then conclude that CoO doping can provide, in agreement with the previously shown permeability and loss results, anisotropy compensation only below cCo = 5000 ppm and temperatures lower than about 60 °C.
IV. CONCLUSIONS
A broadband investigation of permeability and loss behavior of sintered CoO-doped Mn-Zn ferrites has been made as a function of CoO content (up to cCo = 6000 ppm), peak polarization, and temperature. Magnetic softening by CoO addition, namely increase of the initial permeability and decrease of the magnetic losses, is observed to occur at room temperature for cCo < ∼ 5000 ppm. The advantage of doping is observed, however, to weaken beyond a few hundred kHz and with increasing temperatures. It is in any case observed that the resistivity change brought about by doping can interfere with the magnetic losses only at high frequencies, beyond a few MHz. By detailed analysis as a function of frequency, the complex permeability can be decomposed into dw and rotational contributions, with the latter directly linked, via additional knowledge of the saturation magnetization and its dependence on temperature, to the magnetic anisotropy. Assuming then isotropic distribution of the easy axes, the average effective magnetic anisotropy K has been estimated as a function of doping and temperature. K, conventionally taken positive in the host lattice, is always found to decrease monotonically with the temperature, with a trend depending on the CoO concentration. We thus find in the undoped sample K = 36.7 J/m3 at room temperature, decreasing to K = 12.6 J/m3 at T = 120 °C. All the K(T) curves belonging to different cCo values actually tend to converge at such temperature, with minimum and maximum values K = 30.4 J/m3 and K = 64.1 J/m3 associated with cCo = 3000 ppm and cCo = 6000 ppm, respectively, at room temperature. We thus obtain that the contribution KCo(T) from the foreign Co2+ ions subtracts from the anisotropy of the host lattice only below T ∼ 60 °C, crossing the zero value for cCo approaching 5000 ppm and eventually tending to disappear for T > 120 °C.
ACKNOWLEDGMENTS
This work was performed in the context of the Agreement of Scientific and Technological Cooperation between INRIM and CERTH regarding the study of soft ferrites.