Multiphysics simulations of microstructure influence on hysteresis and eddy current losses of electrical steel

This paper employs micromagnetic simulations and computational homogenization on digitized binder-jet printed Fe-Si steel microstructures to demonstrate that optimizing grain size and grain boundary phase thickness can effectively minimize both hysteresis and eddy current losses in electrical steel.

The Gist

Imagine you are trying to build a super-efficient electric motor. The heart of this motor is a special metal called electrical steel. When electricity flows through the motor, this metal acts like a traffic controller for magnetic fields. However, just like a busy highway, the metal isn't perfect. As the magnetic fields switch back and forth, the metal "gets tired" and loses energy as heat. This is called energy loss, and it makes your motor less efficient.

For a long time, scientists have tried to make this metal better by changing its chemical recipe. But recently, a new way of making metal called Additive Manufacturing (basically, 3D printing metal) has opened up a new door. This paper explores what happens inside this 3D-printed metal at a microscopic level and how to make it lose less energy.

Here is the breakdown of their discovery, using some everyday analogies:

1. The Two Enemies: Hysteresis and Eddy Currents

To understand the problem, imagine the metal is a giant crowd of tiny magnets (called magnetic domains) inside a stadium.

• Hysteresis Loss (The "Sticky" Crowd): Imagine the crowd is trying to turn around to face a new direction. Some people are stubborn and stick to their old direction, making it hard to turn the whole group. You have to push really hard (use energy) to get them to flip. This "stickiness" is hysteresis. The paper found that the "glue" between the grains of metal (the grain boundaries) acts like a sticky trap. If the grains are too big, the crowd gets stuck in specific spots, making it harder to turn them around.
• Eddy Current Loss (The "Short Circuit" Crowd): Now imagine the crowd is also running around the stadium track. If the track is a smooth, open loop, they can run fast and easily. But if there are walls or barriers, they have to run in circles or bump into things, creating friction (heat). In the metal, these running paths are electric currents. If the metal is one giant, smooth piece, the currents run wild and create a lot of heat. If you put up walls (insulators) between the grains, the currents get blocked and can't run as far, reducing the heat.

2. The Experiment: Building Digital Twins

The researchers didn't just guess; they built digital twins of the metal.

• They took real 3D-printed metal samples (some with Boron, some without) and took high-powered photos (SEM images).
• They then created two types of computer models:
  1. The "Ideal" Model: They built perfect, computer-generated grains like a mosaic puzzle.
  2. The "Real" Model: They scanned the actual photos of the metal and turned them into a digital map.

They used these maps to simulate how the magnetic crowd behaves and how the electric runners move.

3. The Big Discoveries

By running thousands of simulations, they found some surprising rules about how to tune the metal:

The "Goldilocks" Grain Size

• The Finding: They found that if the grains (the individual "tiles" of the mosaic) are around 120 micrometers wide, the "stickiness" (hysteresis loss) is at its lowest.
• The Catch: However, making the grains bigger makes the "runners" (eddy currents) run faster and lose more energy.
• The Analogy: Think of it like a dance floor. If the floor tiles are too small, the dancers (magnets) trip over the edges constantly. If the tiles are huge, the dancers can spin freely, but the music (electricity) travels too fast and causes a mess. You need a medium-sized tile to keep the dance smooth without the music getting out of control.

The "Thick Wall" Strategy

• The Finding: The space between the grains is filled with a special material (a grain boundary phase). The researchers found that making this "wall" thicker is a win-win.
• The Analogy: Imagine the grains are houses and the boundary is the fence.
  • For Hysteresis: A thicker fence acts like a better buffer zone, helping the magnetic "crowd" switch directions more easily without getting stuck.
  • For Eddy Currents: A thicker fence is a better barrier. It stops the electric "runners" from jumping from house to house. If the fence is thick and resistive, the runners get stuck in their own houses and can't create a big, heat-generating loop.
• Result: Thicker boundaries reduce both types of energy loss.

4. Why This Matters (According to the Paper)

The paper concludes that by simply optimizing the microstructure—specifically by controlling the size of the grains and making the boundaries between them thicker—we can significantly reduce the energy wasted in these magnetic cores.

They proved that you don't necessarily need to invent a new chemical formula; you just need to arrange the existing atoms in a smarter pattern. Their computer models showed that the "thick wall" strategy helps the magnetic material switch directions more easily (less stickiness) while simultaneously blocking the electric currents that cause heat (less short-circuiting).

In a nutshell: The researchers used computer simulations to show that 3D-printed electrical steel works best when the "grains" are a specific medium size and the "fences" between them are thick. This arrangement makes the metal less "sticky" for magnets and better at blocking electric heat, leading to more efficient machines.

Technical Summary

Technical Summary: Microstructural Influence on Hysteresis and Eddy Current Losses of Additively Manufactured Electrical Steel

1. Problem Statement

The efficiency of electrical machines, which rely heavily on soft magnetic materials like Fe-Si alloys, is fundamentally limited by energy losses in the magnetic core. These losses primarily manifest as hysteresis losses (dominant at low frequencies) and eddy current losses. While conventional electrical steels are processed via rolling, high silicon content (>4 wt%) renders the material brittle, complicating traditional manufacturing. Additive manufacturing (AM), specifically binder-jet printing (BJP), offers a novel processing route that can circumvent rolling limitations and has been observed to improve magnetic properties, such as reduced coercivity and losses, particularly when Boron is added. However, the underlying mechanisms linking the specific microstructural features of AM-produced Fe-Si-B steels (e.g., grain size, grain boundary phase morphology) to their magnetic and electrical performance remain unclear. This work aims to elucidate these mechanisms through a multiphysical simulation approach.

2. Methodology

The authors employ a multiphysical framework combining micromagnetic simulations and computational homogenization to analyze digital microstructures.

2.1 Microstructure Reconstruction and Digitization

Two distinct workflows were developed to generate simulation domains:

1. Synthetic Reconstruction: Based on statistical descriptors (average grain size $d_G$, grain boundary thickness $\ell_{GB}$) derived from experimental characterization. This utilizes Poisson disk sampling for seed generation, Voronoi tessellation for grain formation, and the integration of a defined grain boundary (GB) phase.
2. Direct Digitization: Experimental SEM images of two specific samples (Sample A: 3 wt% Si, 0 wt% B; Sample B: 5 wt% Si, 0.25 wt% B) were processed. This involved Gaussian smoothing, threshold-based phase boundary detection, binarization, and connected-component labeling to convert the images into labeled simulation domains.

2.2 Hysteresis Loss Simulation (Landau-Lifshitz)

Hysteresis behavior was simulated using the Landau-Lifshitz (LL) theory implemented in the mumax3 package via Finite Difference Method (FDM).

• Governing Equations: The evolution of magnetization $\mathbf{m}(\mathbf{r})$ is governed by the LL equation, minimizing a free energy functional comprising exchange, anisotropy, magnetostatic, and Zeeman energy terms.
• Material Properties: Magnetic parameters ($K_u$, $M_s$) were defined as functions of Silicon content ($X_{Si}$). The GB-phase was modeled with significantly reduced magnetic properties ($K_u$ and $M_s$ scaled by 0.001) to approximate the mixed Fe-Si/Fe$_2$B lamellar structure.
• Scale Bridging: To ensure numerical validity within the mesoscopic domain (micron scale) while maintaining physical relevance, the Bloch domain wall thickness was artificially set to $\sim 4$ $\mu$m. This required a gradient constant $\kappa_m$ approximately 5,000 times larger than physical values, effectively creating a "model material" that captures the trends of domain wall dynamics (nucleation, pinning, migration) relative to microstructural features.
• Loss Calculation: Hysteresis loss ($P_H$) was derived from the area of the simulated $B-H$ loops over multiple demagnetization cycles.

2.3 Eddy Current Loss Simulation (Magneto-Quasi-Static Homogenization)

Eddy current losses were calculated using a magneto-quasi-static (MQS) approach based on Maxwell's equations, implemented via Finite Element Method (FEM) in the MFM-FM code (MOOSE framework).

• Homogenization: The problem was reduced to a stationary conduction problem to determine the effective conductivity tensor ($\sigma_{eff}$) of the microstructure. This involved solving for electric potentials under linear boundary conditions to satisfy the Hill condition.
• Conductivity Modeling: Grain interior conductivity was derived from experimental data for Fe-Si alloys. The GB-phase conductivity was assumed to be that of bulk Fe$_2$B ($0.018$S m$^{-1}$), significantly lower than the grain interior.
• Loss Calculation: Eddy current loss ($P_E$) was computed using the effective conductivity and the applied AC frequency, considering the skin depth effect. At low frequencies (up to 100 Hz), the loss was approximated using a low-frequency limit formulation.

3. Key Results

3.1 Hysteresis Behavior and Losses

• Role of GB-Phase: The introduction of a GB-phase significantly altered hysteresis behavior. It increased the coercive field ($H_c$) for both samples (e.g., from ~6.8 to ~15.2 kA/m for Sample A) and reduced remanence ($B_r$). The GB-phase acts as a pinning site, restricting domain walls to single grains and promoting nucleation at the boundaries.
• Grain Size ($d_G$) Effects:
  • Coercivity: For 3 wt% Si, $H_c$ showed a slight increasing trend with grain size. For 5 wt% Si, $H_c$ decreased as grain size increased.
  • Remanence: $B_r$ consistently increased with larger grain sizes due to the reduced volume fraction of the low-magnetization GB-phase.
  • Hysteresis Loss ($W_H$): Losses generally decreased with increasing grain size, reaching a minimum at an average grain size of approximately 120 $\mu$m.
• GB-Phase Thickness ($\ell_{GB}$) Effects: Increasing the thickness of the GB-phase led to a reduction in both coercivity and hysteresis loss. This is attributed to the increased volume of the resistive GB-phase which alters the magnetic energy landscape.

3.2 Conductivity and Eddy Current Losses

• Effective Conductivity ($\sigma_{eff}$): The GB-phase significantly reduced the overall effective conductivity of the material. $\sigma_{eff}$ decreased as the GB-phase thickness increased and increased as the grain size increased (reducing the relative volume of the resistive GB-phase).
• Current Flow: In microstructures with resistive GB-phases, electric currents were strongly deflected, flowing primarily through the grain interiors and circumventing the GB-phase. This created "bottlenecks" and preferred pathways.
• Eddy Current Loss ($W_E$):
  • Grain Size: Eddy current losses increased with increasing grain size, as larger grains provide fewer resistive barriers (GBs) to current flow.
  • GB-Phase Thickness: Increasing the GB-phase thickness reduced eddy current losses by lowering the effective conductivity of the bulk material.
  • Silicon Content: Samples with higher Silicon content (5 wt%) exhibited lower losses compared to 3 wt% samples, primarily due to the intrinsic lower conductivity of high-Si Fe-Si alloys.

4. Significance and Claims

The paper claims to provide a validated multiphysical approach for understanding the microstructure-property relationships in additively manufactured electrical steels. Key contributions and claims include:

• Mechanistic Insight: The study demonstrates that the GB-phase is not merely a passive boundary but an active component that influences domain wall pinning, nucleation, and current flow pathways.
• Optimization Potential: The results suggest that microstructure optimization can reduce magnetic core losses. Specifically, the authors identify a trade-off: while larger grains minimize hysteresis loss, they increase eddy current loss. Conversely, increasing the GB-phase thickness helps reduce both hysteresis and eddy current losses, though it may impact saturation flux density.
• Methodological Validation: The close agreement between synthetic reconstructions and SEM-digitized samples suggests that purely synthetic models, based on statistical descriptors, are sufficient for gaining meaningful insights into soft magnetic material behavior, potentially reducing the need for extensive experimental characterization in early design stages.
• Design Guidelines: The work offers specific design guidelines, such as targeting an average grain size of ~120 $\mu$m for minimal hysteresis loss and utilizing thicker GB-phases to mitigate both loss mechanisms, provided the reduction in saturation flux density is acceptable.

The authors maintain a modest tone, acknowledging that the model material uses scaled parameters for numerical stability and that future work is needed to better capture complex features like isolated pores and insular GB-phases, as well as to refine the homogenization of the GB-phase properties themselves.