On the single field formulation in magnetostatics

This paper systematically establishes the equivalence between two variational formulations of magnetostatics—one using magnetization and magnetic field, and the other using only magnetic induction—demonstrating that this link remains stable in coupled magnetoelastic models despite the absence of standard convex duality and the lack of preserved convexity or coercivity in the transformation.

The Gist

Imagine you are trying to describe the behavior of a "smart" material that reacts to magnets, like a piece of rubber that stiffens or bends when you bring a magnet near it. This is called magnetoelasticity.

To understand how this material settles into a stable shape (equilibrium), scientists use math to find the state where the total energy is at its lowest point. This paper tackles a specific puzzle: There are two different ways to write the math for this problem, and the authors want to prove they are actually the same thing.

Here is the breakdown using simple analogies:

The Two Different "Maps"

Think of the material as a landscape. We want to find the deepest valley (the lowest energy state). The paper compares two different maps used to navigate this landscape:

1. The Two-Variable Map (The "Magnetization & Field" approach):

   • This map tracks two things separately: the magnetization (how the tiny magnets inside the material are aligned) and the self-field (the magnetic field the material creates just by being magnetized).
   • Analogy: Imagine trying to describe a crowd of people by tracking exactly where every single person is standing and the wind they create as they move. It's very detailed, but the wind created by one person depends on where everyone else is standing. This makes the math "non-local" and tricky because you have to look at the whole picture at once.

2. The Single-Variable Map (The "Magnetic Induction" approach):

   • This map tracks only one thing: the magnetic induction (the total magnetic effect you can actually measure).
   • Analogy: Instead of tracking every person and their individual winds, you just measure the total wind speed at every point. It's a "local" view—you only need to know what's happening right in front of you to write the equations. This is often easier for computers to solve.

The Big Question

Engineers and physicists have suspected for a long time that these two maps lead to the exact same destination (the same stable shape of the material). However, the paper argues that nobody has rigorously proven exactly when and how this works, especially when the material behaves in complex ways (like being "diamagnetic," which repels magnets, or having "soft saturation," where it can only get so magnetized).

The "Magic Switch" (The Transformation)

The authors show that you can switch between these two maps, but it's not as simple as just swapping one variable for another. You have to use a specific mathematical "magic switch" called the Legendre-Fenchel transform.

• The Catch: This switch only works perfectly if the material's energy rules are "well-behaved" (mathematically, convex or concave).
• The Surprise: The authors found that even though the math for the energy density (the energy in a tiny speck of material) can be transformed using this switch, the total energy of the whole object doesn't always transform nicely in the standard way.
  • Analogy: Imagine you have a recipe for a cake. You can mathematically convert the recipe from "cups of flour" to "grams of flour." But if you try to convert the entire baking process (including the oven heat and the rising time) using the same simple conversion, it might break. The paper proves that for these magnetic materials, the "recipe" conversion works, but the "baking process" (the total energy functional) requires a very careful, specific check to ensure the two maps still agree.

Key Findings in Plain English

1. They are Equivalent at the Finish Line: If you find the stable state (the equilibrium) using the complicated Two-Variable Map, and you translate it to the Single-Variable Map, you get the exact same result. The energy values are identical.
2. They are NOT Equivalent in the Middle: If you pick a random, unstable state (a state that isn't the final equilibrium), the two maps will give you different energy numbers. The "magic switch" only aligns the two maps perfectly when you are standing exactly at the bottom of the valley.
3. Shape Matters: The paper shows that for some materials (like diamagnetic ones that repel magnets), the math looks very different in the two maps. In one map, the energy looks like a bowl (easy to find the bottom); in the other, it looks like a hill (hard to find the top). The authors prove that despite this visual difference, the "bottom of the bowl" and the "top of the hill" correspond to the exact same physical reality.
4. No "Free Lunch" on Convexity: Usually, mathematicians love "convex" problems because they are easy to solve. The paper warns that just because one map is easy (convex) doesn't mean the other map is easy. Sometimes, the easy map is convex, and the other is concave (upside-down). You can't just assume the math behaves nicely in both versions.

The Bottom Line

This paper is a rigorous "proof of concept" for engineers. It says: "You can use the simpler, single-variable math to design these smart materials, and you will get the same correct answer as the complex, two-variable method, provided you use the correct transformation rules and only look at the final stable state."

It clears up confusion in the engineering community by showing exactly where the two methods match and where they diverge, ensuring that when engineers switch between these mathematical models, they aren't accidentally changing the physics of their designs.

Technical Summary

Technical Summary: On the Single Field Formulation in Magnetostatics

Problem Statement
The paper addresses the theoretical equivalence between two distinct variational formulations of magnetostatics, specifically within the context of magnetoelasticity. The first formulation, rooted in Brown's theory, utilizes magnetization density ($m$) and the self-magnetic field ($h_s$) as primary variables, subject to Maxwell's constraints. The second formulation, often preferred in engineering and soft magnetoelasticity (e.g., magnetorheological elastomers), employs a "single field" approach using only the magnetic induction ($b$).

While the engineering literature frequently acknowledges a link between these models via convex duality, the authors argue that the precise nature of this equivalence is often unclear or oversimplified. A critical gap exists in understanding how these formulations relate when:

1. The magnetic energy densities are non-convex or non-concave (e.g., diamagnetic materials or saddle-point problems).
2. The models are coupled with other variational static models, such as elasticity.
3. Hard saturation constraints (e.g., $|m| = m_s$) are present versus soft saturation constraints.

Methodology
The authors employ a rigorous variational analysis to establish the equivalence of the two functionals. The methodology proceeds as follows:

• Formulation Setup: The authors define the total energy functionals for both the $(m, h_s)$-based model ($\tilde{E}$) and the $b$-based model ($E$). They treat the deformation $y$ as a fixed parameter to isolate the magnetic equivalence.
• Duality Transformation: The core of the analysis relies on the Legendre-Fenchel transform (and its generalized version for non-smooth functions). The authors posit that the material energy densities $\Phi$ (for the $b$-model) and $\tilde{\Phi}$ (for the $m$-model) are linked not by a simple identity, but by a specific duality relation involving an augmented energy density $\tilde{\Psi} = \tilde{\Phi} + \frac{\mu_0}{2}|m|^2$.
• Critical Point Analysis: The paper distinguishes between general admissible states and constrained critical points (equilibrium states). The authors analyze the Euler-Lagrange equations for both smooth cases (using gradients) and non-smooth cases (using convex/concave subdifferentials).
• Helmholtz Decomposition: To handle the non-local nature of the self-field in the $(m, h_s)$ formulation, the authors utilize the $L^2$ Helmholtz decomposition on $\mathbb{R}^3$. This allows them to project the magnetization onto curl-free and divergence-free components, effectively linking the stray field $h_s$ to the induction $b$.
• Case Studies: The theoretical results are validated and illustrated through specific examples, including linear diamagnetic/paramagnetic materials, anisotropic mixed materials, permanently magnetized bodies, and soft saturation models.

Key Contributions and Results

1. Equivalence at Equilibrium Only: The primary result is that the two formulations are equivalent only at critical points (equilibrium states). For generic admissible states that do not satisfy the Euler-Lagrange equations, the energy values of the two functionals do not coincide ($E(b) \neq \tilde{E}(m, h_s)$). The transformation linking the variables is an involution only when the system is in equilibrium.
2. Duality Relation: The authors prove that if the energy densities are related by the transformation $\Phi(x, F, b) = -\tilde{\Psi}^*(x, F, b)$ (where $\tilde{\Psi}^*$ is the Legendre-Fenchel transform of the augmented density), then there is a one-to-one correspondence between the constrained critical points of the two models.
3. Non-Convexity and Concavity: The paper demonstrates that the equivalence holds even when the energy densities are not convex.
   • In the case of diamagnetic materials, the $(m, h_s)$ functional is strictly concave and unbounded from below, while the $b$-functional remains strictly convex. The equivalence still holds, but the nature of the critical point changes (minimization vs. maximization).
   • For saddle-point problems (e.g., anisotropic materials with mixed behaviors), the equivalence holds, but the critical point is neither a local minimum nor maximum in the $(m, h_s)$ formulation, despite being a global minimizer in the $b$-formulation.
4. Soft vs. Hard Saturation: The authors clarify the distinction between soft saturation ($|m| \leq m_s$) and hard saturation ($|m| = m_s$).
   • The soft saturation model fits naturally within the convex/concave framework of the paper.
   • The hard saturation model (standard micromagnetics) is not covered by the proposed equivalence because the energy density is neither convex nor concave. The authors show that applying the Legendre-Fenchel transform to the hard saturation model does not recover the original model but rather its convex hull (the soft saturation model). Thus, full equivalence for hard saturation requires relaxation, which is a known result in micromagnetics but is explicitly contextualized here.
5. Energy Equality: It is proven that at the critical points, the energy values of the two functionals are identical ($E(b) = \tilde{E}(m, h_s)$). This equality is a direct consequence of the Fenchel identity and the physical induction relation, holding regardless of whether the densities are smooth or non-smooth, provided the duality conditions are met.

Significance and Claims
The paper claims to fill a gap in the existing literature by providing a precise, rigorous discussion of the equivalence between single-field and multi-field magnetostatic formulations. The authors emphasize that:

• Previous discussions often implicitly assumed convexity or failed to distinguish between equilibrium states and general states.
• The link between the models is not a standard convex duality on the level of the functionals for all states; rather, it is a structural link that ensures the coincidence of energies and critical points specifically at equilibrium.
• The transformation is robust even when coupled with elasticity and when dealing with non-convex or concave magnetic laws, provided the appropriate duality relations are defined.
• The distinction between soft and hard saturation is crucial: the single-field formulation naturally relaxes hard constraints, and the equivalence to the multi-field formulation is only exact for the relaxed (soft) version or when the hard constraint is treated via relaxation techniques.

The authors maintain a modest tone, noting that their results apply to theoretical aspects and smooth or convex/concave settings, and that the full equivalence of non-convex micromagnetic models with hard constraints requires deeper discussion beyond the scope of this specific variational link.