Laser powder bed fusion of a nanocrystalline Finemet Fe-based alloy for soft magnetic applications

Doubling energy efficiency, including the efficiency of electrical engines, has been recently put forward as a critical mandate to restrict the rise in the average temperature of the Earth by 1.5 K,¹ as it would enable a wider decarbonization of transport. A rough (an optimistic) estimate of the energy losses of the 8 × 10⁹ electric motors in the European Union (EU) (Ref. 2) gives the large amount of ∼1350 TWh, which is approximately 30% of the energy consumed annually in a country like Spain. The development of technologies aimed at reducing losses in electric motors is urgent.

Rotors are components of electric motors whose motion is guided by changing magnetic fields.³ Ideally, these components should be made of a material with very soft magnetic behavior, i.e., with low coercivity (H_c), high susceptibility (χ), and high saturation magnetization (M_s), as this would reduce the energy required to develop the desired torque.⁴ Additionally, rotor materials should have high resistivity to limit the eddy current losses at high frequency and high strength to ensure mechanical stability. Traditionally, rotors are composed of laminates of polycrystalline silicon steel,^5,6 a metal that gives rise to relatively high core losses due to the presence of large, elongated grains and of a strong texture. Finding alternative soft magnetic materials and the corresponding manufacturing methods that can shape them into bulk rotor components would, thus, clearly contribute to increasing the overall energy efficiency of electric motors.

Additive manufacturing (AM) methods^7,8 have opened new avenues to produce bulk metallic glass (BMG) components, which were, to date, limited in size because the cooling rates that are achievable in conventional manufacturing are often smaller than those required to freeze the amorphous phase. Laser powder bed fusion (LPBF), for instance, involves cooling rates that may reach up to 10⁸ K/s and, thus, LPBF of BMGs is a rapidly expanding field.^9–13 Most works to date have focused on Zr-based alloys due to their higher glass forming ability (GFA) and to their larger ductility, which makes them more resilient to the thermal residual stresses that are accumulated during processing. Conversely, Fe-based BMGs,¹⁴ which have an excellent soft magnetic behavior, have been less investigated due to their lower GFA and their inherent brittleness, which may even lead to pulverization upon the application of external loads.

Although the study of LPBF of Fe-based BMG alloys is still in its infancy, research has been underway during the past few years to tackle the complex existing challenges.^15–27 The alloys investigated include Fe_54.35Cr_18.47Mn_2.05Mo_13.93W_5.77B_3.22C_0.90Si_1.32 (wt. %),¹⁵ Fe₇₄Mo₄P₁₀C_7.5B_2.5Si₂ (wt. %),¹⁶ Fe_43.7Co_7.3Cr_14.7Mo_12.6C_15.5B_4.3Y_1.9 (at. %),^17,18 FeCrMoBC,^19–22 Fe_68.3C_6.9Si_2.5B_6.7P_8.7Cr_2.3Mo_2.5Al_2.1 (at. %),²³ {(Fe^0.6 Co^0.4)_0.75B_0.2Si_0.05}₉₆Nb₄ (at. %),²⁴ and Fe_73.7Si₁₁B₁₁C₂Cr_2.28 (at. %) (Kuamet 6B2).^25–28 Most of these studies are aimed at relating the LPBF processing parameters [mainly laser power (P) and scan speed (v)] to the resulting structures and to the mechanical and/or magnetic properties of the manufactured amorphous/crystalline composites. It is currently widely recognized that the processing conditions that lead to a high density (high laser power and/or low scan speed) in general result in low fractions of the amorphous phase, and vice versa.¹¹ Thus, achieving simultaneously high values of density and large volume fractions of the glassy phase remains an outstanding challenge in AM of Fe-based BMGs.

Fe-based nanocrystalline alloys such as Finemet (Fe-Si-B-Nb-Cu)²⁹ have also been widely explored over the past few decades as promising soft magnetic materials³⁰ due to their excellent combination of properties, including a large M_s, a high resistivity (double that of the widely used Fe-3 wt. % Si), and an extremely small H_c³¹ (up to an order of magnitude lower than that of silicon steel). Overall, it is recognized that the use of Finemet in electric motors could potentially lead to a drastic reduction of core losses.³² The most common method that has so far been utilized to manufacture nanocrystalline Finemet components is heat treatment of melt-spun glasses,^33,34 as the low GFA of this alloy severely hinders the fabrication of bulk components by casting methods. Alternative techniques include mechanical alloying and milling³⁵ as well as cold sintering and spark plasma sintering,^36,37 which, however, allow a smaller control of crystallization. To date, a limited number of studies have attempted to process Finemet samples using laser-based methods.^38,39 In particular, Conteri et al.³⁸ and Borkar et al.³⁹ report the use of laser engineering net shaping (LENS) to produce FeSiBNbCu alloys from a mixture of elemental iron, silicon, boron, niobium, and copper powders. However, since the cooling rates involved in the LENS process (10⁴–10⁵ K/s) are several orders of magnitude smaller than those required to stabilize the amorphous phase in Finemet (10⁶–10⁸ K/s), the specimens manufactured under all the investigated conditions consist of a high fraction of relatively large dendritic grains and are, thus, endowed with a semihard magnetic behavior. Other studies have attempted nanocrystallization via laser annealing of amorphous ribbons.^40,41 Finally, a single study was found on LPBF of Finemet,⁴² where a continuous laser is utilized to manufacture cylindrical specimens within a processing region comprising laser power values between 60 and 350 W and scan speeds between 250 and 3000 mm/s was explored. The use of LPBF to additively manufacture Finemet alloy parts remains underexplored.

The aim of this work is to explore the feasibility of additively manufacturing the Finemet alloy using pulsed wave LPBF. A processing campaign including variations in the laser powder and the scan speed is carried out in order to establish a relationship between the processing parameters, the (micro)structure, and the magnetic behavior. This work constitutes a pioneer study on the fundamentals of LPBF processing of Finemet alloys aimed ultimately at the production of complex-shaped soft magnetic components with reduced core losses.

LPBF of (8 × 8 × 9.5) mm³ prisms was performed using a pulsed-laser Renishaw AM400 machine furnished with a reduced build volume (RBV) platform containing a 316L steel substrate. A simple meander scan strategy with a rotation of 67° after each layer was utilized in all cases in order to better understand the processing-(micro)structure-property relationships. The hatch distance (h), the point distance (pd), the spot size (ss), and the layer thickness (t) were kept constant in all fabrications. A fixed interpulse delay of 10 μs was utilized. In particular, the values of these parameters were h = 80 μm, pd = 80 μm, ss = 63 μm, and t = 30 μm. Table II summarizes the seven LPBF parameter sets investigated, which include variations of P from 40 to 60 W and of v from 350 to 800 mm/s. The latter was altered by changing the exposure time (t_exp). Higher power values led to a loss of dimensional stability and, thus, have not been utilized in the present study. Finally, the equivalent volumetric energy density (VED) values were also calculated and are listed in Table II next to the corresponding processing parameter set.

The defect structure of the LPBF-manufactured prisms was first examined by optical microscopy (OM) using an Olympus BX51 microscope and a wide area of the sample was imaged by stitching together micrographs captured at a magnification of 5×. The software imagej was then utilized to classify and to separate different defect types (pores and cracks) from the bulk. In particular, Fiji’s machine learning-assisted segmentation plug-in, the WEKA trainable segmentation package,⁴³ was used for this purpose. Samples for OM examination were first cut using a Secotom disc cutting machine, and they were then ground sequentially with 320, 600, 1200, and 2000 grit SiC papers. Finally, surface finishing was conducted by mechanical polishing with diamond suspensions of average particle sizes amounting, consecutively, to 6, 3, and 1 μm. The phase distribution and the fraction of the amorphous phase in the LPBF-manufactured specimens were characterized, respectively, by XRD and by DSC following a similar procedure to that utilized for the powders, which is explained above. The sample preparation procedure for XRD characterization consisted of the same cutting, grinding, and mechanical polishing steps described above for the preparation of sample surfaces for OM examination. Sample preparation for DSC measurements included cutting with a slow speed diamond saw specimens of approximately 20 mg in weight. The microstructure (grain size, shape, and texture) of the crystalline regions was examined by electron backscattered diffraction (EBSD) using an Apreo 2S (ThermoFisher Scientific) field emission gun scanning electron microscope (FEGSEM) equipped with an Oxford Instruments detector, a CCD camera, and the Aztec data acquisition and analysis software package. The measuring conditions included an accelerating voltage of 20 kV and a step size of 70 nm.

Room-temperature magnetization behavior of the LPBF-manufactured specimens was studied by means of a vibrating sample magnetometer (VSM, Lakeshore) using a continuous measurement procedure in the applied field range ±1300 kA/m. Hysteresis loops were measured at room temperature and, from them, M_s and H_c were estimated for each sample after appropriate subtraction of the diamagnetic signal from the sample holder.

Figure 1(a) is an SEM secondary electron micrograph illustrating the morphology of the Finemet feedstock powder. The particles have ellipsoidal to spherical shape, with a small fraction of satellites. The size distribution, plotted in Fig. 1(b), is monomodal and it is characterized by D₁₀, D₅₀, and D₉₀ values of 18, 35, and 58 μm, respectively. The Hall flowability of the powder is 18 s, and D_a and D_t amount, respectively, to 3.9 and 4.3 g/cm³. The compressibility index and the Hausner ratio, calculated using Eqs. (1) and (2), are 1.08 and 7.87, respectively. These values are considered indicators of an “excellent” flow character.⁴⁴ 

Figure 2 shows, on the one hand, the LPBF-manufactured prisms that were processed using the parameter combinations 1–7 that are listed in Table II. It can be seen that a good dimensional accuracy is achieved for all the printing conditions. Optical micrographs of a cross section parallel to BD that depicts the defect structure of all the Finemet printed prisms are also displayed in Fig. 2. The density corresponding to each sample, measured by image analysis from the optical micrographs, is indicated in the top right corner of the corresponding image. It can be seen that the part density increases with increasing P and with decreasing v, as generally observed in metallic glasses that are manufactured in the conduction mode,¹¹ as well as for other alloys. Table III lists the density values corresponding to the seven LPBF-manufactured samples, which range from 70% to 89%. This table also shows the area fraction of cracks, which remains in general below ∼2%, and the fraction of lack of fusion pores, which varies between ∼11% and ∼30%. In order to test the robustness of the LPBF process, six additional prisms (R1–R6) were manufactured with the conditions of sample 5 (60 W, 500 mm/s) (Supplementary Table 1).⁴⁷ The defect structure of these samples is shown by means of the corresponding optical micrographs along a plane parallel to BD in Supplementary Fig. 1,⁴⁷ and the corresponding values of the density and the fraction of pores and of cracks are included in Supplementary Table 2.⁴⁷ The average value of the density of sample 5 plus the R1–R6 samples amounts to 88.0% ± 0.9%. The small error obtained attests to the robustness of the study. An additional set of prisms with the conditions corresponding to samples 1–4, termed S1–S4, were also manufactured and the corresponding data are also included in Supplementary Tables 1 and 2,⁴⁷ as well as in Supplementary Fig. 1.⁴⁷ Again, the similarity in the values obtained attest to the reproducibility of the study.

Figure 3 compares the XRD diagrams corresponding to all the Finemet LPBF-manufactured parts with that of the atomized powders and of a melt-spun ribbon of the same composition. The powders and the ribbon are fully amorphous, as indicated by the sole presence of a wide hump at 45° in their XRD patterns. Irrespective of the processing conditions, the patterns of the LPBF-manufactured prisms consist of a superposition of a wide hump located around 45° and sharp peaks corresponding to the phases α-Fe/Fe₃Si, which cannot be distinguished by XRD, and to the intermetallic Fe₂B₆. That is, XRD measurements confirm, qualitatively, the presence of an amorphous/crystalline composite in all the printed parts. Figure 4 depicts the variation of the volume fraction of the amorphous phase, measured by DSC, with the laser power and with the scan speed. Within the range of processing parameters investigated, the fraction of the glassy phase varies from 13% to 49%. Figure 4 evidences a strong influence of the scan speed on the fraction of amorphous material, which increases significantly with increasing v. Comparison of Figs. 2 and 4 reveals that the processing parameter windows that give rise to the highest densities and to the highest amorphous fractions, respectively, do not overlap. Figure 5 shows the variation of the relative density with respect to the amorphous fraction for all the LPBF-manufactured samples. Even though the data presented is limited, it can be seen that the highest AM% is, indeed, obtained in the samples with the lowest density.

The microstructure of the crystalline regions was found to be fairly independent of the processing conditions. Figure 6 illustrates two EBSD inverse pole figure maps in the BD captured in representative crystalline regions of samples 5 (60 W, 500 mm/s, density = 89%, AM% = 27%) [Fig. 6(a)] and 1 (40 W, 350 mm/s, density = 81%, AM% = 13%) [Fig. 6(b)]. It can be seen that, in both cases, the microstructure is formed by fine and equiaxed crystallites that are randomly oriented. Figure 6(c) shows, additionally, that the distribution of grain sizes, measured as the equivalent grain diameter by EBSD, is also similar in both cases. For both samples, crystallite sizes lie mostly between ∼250 nm and 2 μm, with average values (⟨d⟩) of 814 ± 386 nm for sample 5 and 710 ± 395 nm for sample 1. It must be taken into account that under the current observation conditions EBSD might not be able to detect the smallest grains and, therefore, the presence of grains with diameters smaller than 250 nm cannot be completely ruled out. Nevertheless, as can be seen in Figs. 6(a) and 6(b), the fraction of indexed grains is relatively high.

Room-temperature hysteresis loops of all Finemet LPBF-manufactured samples are reported in Fig. 7. In particular, the curves of specimens processed with laser power values of 40 [Fig. 7(a)] and 60 W [Fig. 7(b)] are shown in separate graphs. In the inset, a low-field magnification is shown to evidence the coercivity values. The area of the plot close to the origin is depicted at higher magnification as an inset. The magnetization behavior of a fully amorphous melt-spun ribbon is also included in the plots for comparison. In general, in comparison to the ribbon, the LPBF-manufactured samples exhibit similar saturation magnetization values, reached at higher fields with respect to the amorphous ribbons, lower permeability, and higher coercivity. The coercive field value of the ribbon (13 A/m), being below the sensitivity of the VSM, was measured by using a digital wattmeter at 5 Hz exploiting a conventional fluxmetric technique. Figure 8 illustrates the variation of the coercive field as a function of the LPBF processing parameters and, in particular, of the laser power (P) and of the scan speed (v). For a constant v, different P values do not affect H_c substantially, although a minor decrease in H_c is observed with increasing P. Conversely, by increasing v between 350 and 800 mm/s, minor changes in H_c can be observed. In particular, for P = 40 W, the lowest H_c values are obtained at intermediate scan speeds of 500 and 650 mm/s. For P = 60 W, H_c decreases slightly when increasing v from 500 to 800 mm/s. The H_c values corresponding to the processing conditions investigated in this work are summarized in Table III. H_c values range in the interval from 1180 to 1450 A/m. The average value of H_c in sample 5 plus the R1–R6 samples (see Supplementary Table 2)⁴⁷ amounts to 1328 ± 26 A/m. The small error obtained attests to the robustness of the study. The H_c values corresponding to the S1–S4 prisms manufactured with the conditions corresponding to samples 1–4 are also included in Supplementary Table 2.⁴⁷ Again, the similarity in the values obtained attests to the reproducibility of the study.

Figure 9 illustrates the relationship between H_c and two key microstructure features such as the sample density and the amorphous fraction. H_c turns out to be substantially independent of the density. Concerning the amorphous fraction, a slight dependence is observed. In particular, an increase in the fraction of the amorphous phase from 13% to 49% (i.e., 3.8 times) leads to a decrease in H_c of about 18%. This behavior turned out to be different from what is observed in amorphous Fe-based metallic glasses such as Kuamet 6B2,⁴⁵ where the coercive field halves when the amorphous fraction increases by a comparative amount.

The existence of a strong dependence of H_c on the average grain size (⟨d⟩) in nanocrystalline soft magnetic materials is well known in the literature.^32,46 Figure 10 plots the dependence of the coercive field on ⟨d⟩ of soft magnetic materials in the whole ranges of structural correlation lengths (red dotted line).³² In the ultrafine grain regime (⟨d⟩ ≲ 80 nm), H_c is proportional to ⟨d⟩.⁶ However, when ⟨d⟩ ≳ 0.1–1 μm,³²  H_c decreases rapidly with increasing grain size. The blue dots in Fig. 10 indicate the (⟨d⟩, H_c) data corresponding to the LPBF-manufactured sample 5, with the highest H_c (1450 A/m), and sample 1, with a lower H_c (1260 A/m). Both data points fit well with the overall trend depicted by the dashed red line, suggesting that the grain size is a key contributor to the coercivity. The small variations of H_c found in the seven LPBF-manufactured samples would be consistent with the presence of similar ⟨d⟩ values. This study suggests that reducing the coercivity to levels that are comparable to those of melt-spun ribbons (13 A/m) would require the stabilization of crystalline microstructures with average grain sizes either close to 10 nm or of the order of several hundred microns.

The aim of this work was, first, to investigate the feasibility of manufacturing the Finemet nanocrystalline soft-magnetic alloy via pulsed wave laser powder bed fusion. With that goal, simple geometry specimens were produced using a range of laser power and scan speed combinations and the processing parameters were subsequently related to the (micro)structure and to the magnetic behavior. The following conclusions can be drawn from the present study:

1. Finemet prisms with dimensions 8 × 8 × 9.5 mm³ were manufactured with good dimensional accuracy within a processing window including laser power values ranging from 40 to 60 W and scan speeds comprised in a 350–800 mm/s interval. The relative density increased with increasing P and decreasing v, while the amorphous fraction was directly proportional to v but almost independent of P. Densities as high as 89% and amorphous fractions as large as 49% were obtained within the investigated processing window. The microstructure of the crystalline regions was found to be formed by ultrafine, equiaxed grains with random orientations.

2. In comparison to fully amorphous melt-spun Finemet ribbons, the LPBF-processed samples exhibit a similar saturation magnetization, lower permeability, and higher coercivity, irrespective of the processing conditions.

3. The coercive field of the additively manufactured specimens is fairly independent of the relative density and exhibits a moderate inverse variation with the amorphous fraction. The average grain size appears to be a key contributor to H_c.

.

This work has been carried out under the scope of the AM2SoftMag project, funded by the European Innovation Council through the HORIZON-EIC-2021-PATHFINDEROPEN-01 grant (GA No. 101046870). The authors thank the rest of the AM2SoftMag partners for their involvement in the project. Extended acknowledgment is given to Amalia San Román and Ignacio Escobar for their assistance with LPBF trials, to Javier García Molleja for his support with XRD measurements, and to Maryam Rahimi Chegeni for her upholding in calorimetry measurements.

The authors have no conflicts to disclose.

S. Sadanand: Investigation (lead); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). M. Rodríguez-Sánchez: Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). A. Ghavimi: Investigation (equal); Methodology (equal); Writing – review & editing (equal). R. Busch: Funding acquisition (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). P. Sharangi: Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). P. M. Tiberto: Funding acquisition (equal); Resources (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). E. Ferrara: Investigation (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). G. Barrera: Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). L. Thorsson: Funding acquisition (equal); Methodology (equal); Writing – review & editing (equal). H. J. Wachter: Funding acquisition (equal); Methodology (equal); Writing – review & editing (equal). I. Gallino: Funding acquisition (lead); Resources (equal); Supervision (equal); Writing – review & editing (equal). M. T. Pérez-Prado: Funding acquisition (equal); Resources (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

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