Melt-spun Fe-based ribbons are widely used as the core of transformers and inductors due to their high flux density and low coercivity. However due to their high thickness (∼19 µm) these ribbons are prone to large eddy current losses at MHz frequencies. Despite low yield, ball milling has been widely used to break such ribbons down to thinner flakes to suppress the eddy current losses at high frequency. In this work, we demonstrated an optimized ball milling process with increased yield for flakes in the desired size range (2–4 µm). We have demonstrated that reducing pre milling annealing temperature from 450 to 350 °C increases the yield in desired size range from 2% to 5% and further increasing batch size from 10 to 20 g increases the yield to 21%. The coercivity of the milled flakes increases from 139 to 1352 A/m due to the ball milling process. A post-mill annealing at 350 °C in Ar atmosphere decreases the coercivity to 341 A/m. X-ray diffraction analysis showed no sign of crystallization during ball milling. The result presented here demonstrates an efficient approach to fabricate ultra-thin flakes out of soft magnetic ribbons for high-frequency applications.
In the past years, industry has been moving towards the miniaturization of electronic components to enable higher performance and efficiency. Passive devices, including power inductors and transformers, typically have a large footprint on a circuit board. By increasing the switching frequency f of the power management integrated circuit, the required inductance value can be reduced, which in turn, reduces the inductor footprint. Therefore, the industry has been witnessing an increase in the power conversion frequency from kHz to the tens of MHz and its continual growth towards hundreds of MHz. However, such an increase in frequency of operation demands reduced core losses to maintain the inductors’ high efficiency.
It is reported that about 9% of electrical energy generated is lost as heat dissipation during electromagnetic transmission and distribution.1 Typical core losses consist of three major parts: hysteresis, eddy current, and anomalous losses. The hysteresis loss is proportional to the switching frequency f, while the eddy current loss is proportional to f.2 Therefore, at higher frequencies, eddy current loss becomes the dominant part of the total losses, making them the focal points for developing new materials for inductor cores. Increasing the resistivity of the core and minimising the dimension of the magnetic filler, which is perpendicular to the magnetic flux lines, are two strategies to mitigate the eddy current loss. In this regard, soft magnetic composites (SMCs) hold promise for use in power electronics owing to their flexibility in the core shape, their high flux density, and low losses.2–4
To improve the performance of the SMCs, insulated magnetic flakes can be used as the magnetic filler. Compared to spherical magnetic particles, magnetic flakes are beneficial due to their lower demagnetizing factor.5 In addition, the high resistivity of amorphous or nanocrystalline materials are superior magnetic fillers compared to their low-resistivity crystalline counterparts. Gas atomisation process, known to produce spherical amorphous powder, suffers from low quenching rate that makes the large-scale amorphization of Fe-based powder more challenging.5,6 In this regard, amorphous ribbons fabricated by melt-spinning processes are more suitable owing to their very high cooling rate and enhanced amorphization. Such melt-spun ribbons offer excellent DC and low-frequency magnetic properties,7,8 but their thickness in the order of tens of µm does not satisfy the pre-requisites of high-frequency applications. Therefore, the ribbons need to be further thinned and broken into small pieces to be applicable as SMC fillers. The as-quenched ribbons are hard to break because their amorphous structure7 offers high ductility and hardness. Therefore, partial crystallization of such ribbons through annealing and subsequent high-energy ball milling can transform them into thin flakes, ideal for high-frequency SMCs.9 An optimal ball milling condition is in demand.
Ball milling as a stochastic process involves several effective variables. For instance, the physical and chemical properties of the ball milling product can be different depending on the milling medium, rotational speed, number and size of balls, ball-to-powder weight ratio (BPR), and milling time.10 High-energy ball milling induces stresses, which can be successfully reduced by annealing the resulting powders at temperatures below the crystallization temperature of the amorphous powder. The formation of magnetic oxide phases during milling on the reactive surface of the powder particles or pinned domain walls by the rough surface of melt-spun flakes is known to increase the coercivity of the milled material. Recently, flake-like particles from a variety of precursors (FeSiAl,11 FeSi,12 ZrCu,13 Fe, and FeNi14) have been reported. Efficient thinning of the ribbons usually takes place when ball milling is done in the wet condition using an organic liquid as the milling medium. This minimizes the cold-welding mechanism. In addition to efficient thinning, the efficiency of the ball milling procedure, as a large-scale production method, in forming the particles with a high aspect ratio is crucial for the fabrication of SMCs with high permeability and low loss. However, finding an efficient ball milling recipe through which ultra-thin ribbons with a high aspect ratio can be produced has been overlooked.
Here, we demonstrate an optimized high-energy ball milling recipe in ethanol medium to thin down the amorphous Fe-based ferromagnetic ribbons into tiny (thickness:2–4 µm, sizes: 40–60 µm) flakes. Further, post-mill anneal step is used to keep the coercivity of the milled flakes low. In addition, BLENDER simulation is also used to identify the desired flake size range. To the best of our knowledge, this is the first study on the optimization of the ball milling variables to increase the production yield of particles with high aspect ratios such as flakes.
II. EXPERIMENTAL DETAILS
A. Sample preparation
The starting material is Fe83.21Cu1.22Nb5.44Si8.69B1.44 (at. %) as-quenched ribbon with a thickness of 19 µm and width of 8 mm (manufactured by Vacuumschmelze). According to the manufacturer’s datasheet, the ribbon has a crystallization temperature of 510 °C. To embrittle the ribbons and hence more susceptible to ball milling, they were annealed in an Ar atmosphere at 350 °C or 450 °C for 60 min as shown in Table I. Annealing was in Ar to avoid oxidation of the ribbons. They were then cut into small pieces of about 1–2 cm in length and milled. Ball milling was done on a Retsch PM100 planetary miller for 10 h in a Zirconium oxide jar with a volume of 250 ml. The milling media was Zirconium balls with diameters of 20 mm. The ball-to-powder ratio (BPR) was fixed at 15, and the rpm was fixed at 200. Ethanol was used as a milling medium. The produced powders were then removed from the jar, placed in a Petri dish, washed in ethanol, and dried on a hotplate set to a temperature of 40 °C. After the ethanol fully evaporated, the powders were then separated into size ranges of 0–40, 40–63, 63–106, and >106 µm by mechanical sieving. The amount of powder in each size range was then weighed to determine the yield. Furthermore, post-mill annealing at 350 °C for 1 h in Ar is carried out for a selected batch of flakes with sizes of 40–63 µm to keep the coercivity low.
|.||Milling conditions .||Results .|
|Annealing temperature (°C) .||Ball milling time (h) .||Batch size (g) .||Flake thickness (µm) .||Yield in 40–63 µm range (%) .|
|.||Milling conditions .||Results .|
|Annealing temperature (°C) .||Ball milling time (h) .||Batch size (g) .||Flake thickness (µm) .||Yield in 40–63 µm range (%) .|
Magnetometry study was conducted on both pre-milled and post-milled samples using an SHB Mesa 200 HF. Ribbons samples were cut into 20 mm long piece and placed at the centre of the sample holder. Powder samples were attached to a Scotch™ tape (20 × 8 mm) and taped down at the centre of the sample holder for measurements.
The morphology and particle sizes of the samples were investigated by scanning electron microscopy (SEM, Zeiss Supra 40). X-ray diffraction (XRD, Phillips Xpert diffractometer, Cu Kα = 1.54 Å) was used to determine the amount of crystallization in these amorphous alloy samples during the milling stage.
III. RESULTS AND DISCUSSION
The target flake size range for the soft magnetic composite with high permeability and low power loss is 40–63 µm with a thickness of 2–8 µm (more discussion on this below). The low thickness enables significant suppression of eddy currents while the relatively larger flake size allows a higher permeability in the fabricated cores.
We first ball milled the ribbons at different milling times to determine how the ball milling progressed with increasing milling time as tabulated in Table I. Therefore, after the milling time optimization, 10 g of ribbons were annealed at 450 °C and then ball milled for 10 h. We selected the 10-h milling time as the thicknesses of the milled flakes were not sufficiently thin with a shorter milling time. SEM was done on the resulting flakes as well as the starting ribbon materials as shown in Fig. 1. The initial thickness of the ribbons is 19 µm [Fig. 1(a)], corresponding to the manufacturer’s specification, while, after 10 h of ball-milling, the resulting flake powder become thin (2–4 µm) as shown in Fig. 1(b). A shorter milling time is not enough to reduce the flake thickness as identified in Figs. 1(c) and 1(d).
It is to be noted that the milled flakes have a large size distribution and relatively narrow thickness distribution. To understand the flake size range required to achieve high permeability, and thus high packing density during the formation of a composite core with these flakes as the magnetic inclusion, we carried out a simulation in Blender to study the flake size vs packing density. To simplify the simulation, circular disks with varying diameters were used in the simulation to mimic the flakes. The thickness of the disks was fixed in the range between 2–4 µm. Disks with different sizes were used to fill in a box container with the size of 1000 × 1000 × 300 µm3 [Fig. 2(a)]. We found that the disks with a diameter range of 40–60 µm resulted in the best packing density of 59.7%. Disks with smaller diameters between 20–50 µm exhibited a packing density of 32.5%. Whereas disks with the larger diameters of 60–100, 100–130, and 130–150 µm diameter showed a packing density of 48.9%, 37.3%, and 33.2%. Based on our observation from the Blender simulation, we decided to target flakes in the size range close to 40–60 µm in this study.
The powder milled for 10 h was then sieved with mesh sizes of 106, 63, and 40 µm to separate the flakes into different sizes. Subsequently, the weight of powder in each size range was measured, and the ball milling yield calculated. The yield for >106, 63–106, 40–63, and and <40 µm are 0%, 4%, 3%, and 92%, respectively. The low yield in the desired size range of 40–63 µm makes it very challenging to prepare enough powders to fabricate inductor cores. Hence it was necessary to optimize the conditions and increase the yield in such a size range. Accordingly, we changed the ball milling conditions including the pre-mill annealing temperatures of the ribbons. Table I summarizes the conditions we varied and the related yields from this optimization. We first reduced the annealing temperature to 350 °C. The results are listed as Sample 5 in Table I. The 5% yield achieved is a noticeable improvement over Sample 1 that was annealed at 450 °C. In parallel to reducing the temperature, we also increased the batch size from 10 to 20 g. The results are shown as Sample 6. The yield is found to increased from 3% to 12% (Table I) while the flake thickness is constant between 2 and 3.5 µm. Lastly, with Sample 7, we combined lower annealing temperature with a larger batch size and milled 20 g of ribbons annealed at 350 °C. The yield was measured to be 21%, having a flake thickness of 2–8 µm – a significant increase of overall yield for the desired size range of 40–63 µm. The ribbons annealed at 350 °C are more ductile, so it is more likely to produce larger flake sizes by milling them for 10 h, compared to the ribbons annealed at 450 °C. In addition, increasing the batch size possibly enhance the chance of ball-to-ball and ball-to-container wall collisions in which there are multiple parallel aligned flakes. This results in the simultaneous thinning of flakes without extensive fracturing, leading to higher milling yield for larger particles.
The magnetic properties of the pre-milled ribbons, flakes, and annealed flakes were studied to estimate the effect of milling and annealing on the coercivity of these materials. The measurement results are shown in Fig. 3 below. First, the annealed pre-milled ribbon shows a coercivity (Hc) of 122 A/m (green curve). After 10 h of ball milling, due to induced stress and microstrain into the atomic structure, the Hc increases to 1352 A/m (red curve). Such high Hc would contribute to higher losses at low and high frequencies, and thus not desirable. It is found that post-mill annealing can partially restore the low coercivity. The flakes were then annealed at 350 °C for one hour in Ar atmosphere, which resulted in a significant reduction in the coercivity, Hc = 341 A/m (blue curve).
Eddy current loss and hysteresis loss are two major losses in inductor core. The eddy current loss per unit volume is given by the following equation:
where Ke = π2/6ρ is eddy current coefficient and depends on the material resistivity ρ. Bmax is the peak magnetic flux density in the core and it depends on the core design, maximum current going through the winding, and permeability of the core material. f is the frequency and t is the core thickness or laminate thickness, in our case this is the flake thickness. Hysteresis loss per unit volume, on the other hand, is expressed by:
η and n are Steinmetz hysteresis coefficients. η is approximately proportional to Hc of the core material. n in most cases has a number between 1.5 to 2.5. For high frequency applications, as f increases, eddy current loss Pe can become the dominant component of the core losses because Ph is proportional to frequency and Pe is proportional to the square of the frequency. The equation for Pe also shows that reducing flake thickness t can effectively suppress the Pe because it is also proportional to the square of t. Our fabricated flakes have a reduced thickness (2–8 µm) compared to the original ribbons (19 µm). Hence it has the potential to substantially reduce the eddy current loss Pe. We note that our flakes have higher Hc compared to the starting ribbons. However, as discussed above, this will increase the hysteresis loss slightly, which is not a major contributor to the overall losses in the core at a high frequency application.
The formation of a large crystalline phase is detrimental to the overall magnetic behaviour of these amorphous ferromagnetic flakes. Annealing, albeit at a sub-crystallization temperature, was done both during the pre-mill and post-mill stages. Therefore, it is necessary to investigate the nature and extent of the crystalline phase in the annealed flakes, if any. XRD was thus performed on the as-received ribbons, annealed ribbons, milled flakes, and milled/annealed flakes (Fig. 4). The broad peak in the range of 2ϑ = 40°–50° confirms the amorphous nature of the as received, annealed and ball milled samples. After post-mill annealing at 350 °C for an hour, a tiny and broad peak appears in the range of 40° < 2ϑ < 50° that can be assigned as body centered cube (bcc) Fe (110).15 This indicates that, although the annealing temperature is well below the crystallization temperature of the material (510 °C), nano-crystallization occurred during the post-mill annealing. The sharp crystallization of the original ribbons at 540 °C (shown in the inset of Fig. 4) compared to the peak in annealed (350 °C) flakes attests to the sporadic nano-crystalline nature of the flakes. Moreover, crystal size estimated using Scherrer’s equation (D = ) shows, the nanocrystals are in the order of ∼11 nm. β is the full width at half maximum in radian and θ is the Bragg angle corresponding to the (110) peak. The formation of nanocrystallites can possibly be due to the formation of activated flake surfaces because of severe mechanical deformation during ball milling that are prone to oxidation. However, having Cu in the alloy composition helps to have more nucleation sites in the alloy, while having Nb helps to pin the grain boundaries. Together, both these effects can hinder the grain coarsening, and thus, the resulting powder exhibits relatively low Hc based on the random anisotropy model.
We demonstrated a systematic development of a ball milling recipe for preparing high aspect ratio flakes out of thick amorphous ferromagnetic ribbons with high yield and without compromising on the coercivity of the material. We reduced the ribbons’ thickness from 19 µm to 2–4 µm. We also increased the ball milling yield for the larger particles (size range 40–63 µm) from 3% to 21% by decreasing the pre-mill annealing temperature from 450 to 350 °C and by increasing the batch size from 10 g to 20 g. The yield of 21% is significant considering the large flake dimensions. This makes it more efficient and industrially feasible to produce thinned magnetic flake powders. We also studied DC magnetic properties of the as-quenched ribbons and processed powders. Inducing stress and microstrain in the atomic structure through ball milling increases the coercivity from 139 A/m to 1352 A/m. Nevertheless, this was reduced to 341 A/m by performing post-mill annealing. Our XRD measurements indicate no crystallization until post-mill annealing. Although nano-crystallization is observed in the powders being annealed after ball milling, it does not impact the performance of the material negatively. The proposed procedure for soft magnetic flakes fabrication with optimized yield can play a major role in large-scale production of soft magnetic composites which are the main components of efficient inductors and transformers for power conversion, especially at a higher frequency such as beyond 5 MHz.
We would like to thank Amber/Science Foundation Ireland for funding of research-Amber RC12578-P2, SFI-12/RC/2278_P2-Amber-TP01.
Conflict of Interest
The authors have no conflicts to disclose.
Liang Ye: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Writing – original draft (lead); Writing – review & editing (lead). Rajasree Das: Data curation (supporting); Formal analysis (supporting); Methodology (equal); Writing – review & editing (equal). Guannan Wei: Formal analysis (equal); Methodology (supporting); Writing – review & editing (supporting). Sumit Sukhbasi Lal: Methodology (supporting); Writing – review & editing (supporting). Michael Morris: Supervision (supporting); Writing – review & editing (supporting). Hasan Ahmadian Baghbaderani: Resources (equal); Supervision (equal); Writing – review & editing (supporting). Ranajit Sai: Resources (equal); Supervision (equal); Writing – review & editing (equal). Paul McCloskey: Resources (equal); Supervision (lead); Writing – review & editing (equal).
The data that support the findings of this study are available from the corresponding author upon reasonable request.