An energy-decreasing algorithm for the finite element approximation of ferronematic equilibrium states

Technical Summary: An Energy-Decreasing Algorithm for the Finite Element Approximation of Ferronematic Equilibrium States

Problem Formulation
This work addresses the numerical approximation of two-dimensional ferronematic equilibrium states. Ferronematics are composite systems consisting of magnetic nanoparticles dispersed in a nematic liquid crystal. The equilibrium states are modeled as minimizers of a free energy density. In the specific asymptotic limit considered here, where the ratio between the material-dependent length scale and the computational domain length scale is very small, the problem reduces to minimizing the harmonic energy of two coupled vector fields: the nematic order parameter $\mathbf{Q}$ and the magnetization vector $\mathbf{M}$.

Both fields are subject to prescribed length constraints and a nonlinear orientation relation. Specifically, the state vector $\Psi = (\mathbf{Q}, \mathbf{M})^T$ is constrained to lie in a connected, compact, one-dimensional smooth submanifold $\mathcal{N}$ of $\mathbb{R}^4$. The mathematical problem is formulated as minimizing the harmonic energy functional:
$$E(\Psi) := \frac{1}{2} \int_{\Omega} \|\nabla \Psi\|^2_{\mathbb{R}^4 \times 2} \, dx$$
subject to $\Psi(\mathbf{x}) \in \mathcal{N}$ almost everywhere in the domain $\Omega$, with prescribed Dirichlet boundary conditions compatible with the constraint.

Methodology
The authors propose a finite element approximation using piecewise affine, continuous elements on a weakly acute triangulation. A key aspect of the methodology is the handling of the non-convex constraint manifold $\mathcal{N}$.

1. Continuous Level Algorithm: The authors first devise an energy-decreasing iterative algorithm at the continuous level. This relies on a representation of the manifold $\mathcal{N}$ via a unit vector $\mathbf{n}$ from the unit circle $\mathbb{S}^1$. The algorithm involves projecting a linear update onto the sphere to ensure the constraint is satisfied. Theoretical analysis proves that this sequence of updates strictly decreases the energy functional.

2. Discrete Level Algorithm: Extending this to the discrete level is non-trivial because the continuous identity relating the gradients of $\Psi$ and $\mathbf{n}$ does not hold for discrete fields. The authors introduce two key innovations:

   • Weakly Acute Mesh Property: They exploit the weakly acute property of the triangulation to express the discrete harmonic energy as a sum of local positive contributions indexed by pairs of mesh nodes.
   • Geometric Lemma: A crucial geometric property of the manifold $\mathcal{N}$ (Lemma 3.2) is utilized to relate the updates of the two unit vectors associated with the nodes. This allows the construction of a discrete energy-decreasing iteration.

3. Computational Realization: To solve the resulting constrained minimization problem efficiently, the authors employ a decomposition-coordination framework based on an augmented Lagrangian approach.

   • The nonlinear constraint is decoupled by introducing an auxiliary variable and a Lagrange multiplier.
   • A Uzawa-like iterative method (specifically, an Alternating Direction Method of Multipliers or ADMM) is used to find the saddle point of the augmented Lagrangian.
   • To enhance computational efficiency, the nonlinear coupling function in the primal update step is linearized at each iteration, converting the sub-problem into a linear system solvable via direct methods (QR decomposition).

Key Contributions

• Algorithm Development: The paper presents a novel energy-decreasing iterative algorithm specifically tailored for the ferronematic equilibrium problem under length constraints.
• Theoretical Guarantees: The authors provide a rigorous proof that the proposed discrete algorithm guarantees a non-increasing sequence of energy values, provided the triangulation is weakly acute. This is achieved through a specific geometric lemma regarding the constraint manifold.
• Nonconforming Approximation: The method utilizes a nonconforming approximation where the constraint is enforced strictly at the mesh nodes, while the discrete functions take values in the manifold only at these nodes.
• Efficient Solver: The integration of the decomposition-coordination strategy with a linearized Uzawa/ADMM scheme offers a computationally viable approach to solving the coupled nonlinear system.

Numerical Results
The algorithm is tested on two examples involving smooth Dirichlet boundary conditions on a square domain, with known analytical solutions.

• Convergence: Numerical experiments demonstrate that the energy error decreases as the mesh is refined. For the first test case, first-order convergence is observed in both the $H^1$-seminorm and $L^2$-norm. For the second, more complex test case, first-order convergence is observed in the $H^1$-seminorm, while second-order convergence is achieved in the $L^2$-norm.
• Parameter Sensitivity: The study analyzes the influence of the augmentation parameter ($\zeta$), the update parameter ($\rho$), and the Uzawa tolerance ($\varepsilon_{pri}$). Results indicate that while larger $\zeta$ values increase the iteration count significantly, they do not necessarily improve the accuracy-to-cost ratio. Tighter tolerances are required on finer meshes to prevent oscillations in the energy decay of the outer loop.
• Performance: The method successfully converges to equilibrium states with controlled energy errors, validating the theoretical energy-decreasing property.

Significance
The paper claims to provide the first numerical analysis and algorithmic framework specifically for the ferronematic equilibrium problem in the asymptotic limit described. While the minimization of harmonic energy under length constraints is well-explored for simpler models (such as the Oseen–Frank model or the Eriksen model), this work extends these concepts to the coupled ferronematic system. The primary contribution is the development of a robust, energy-decreasing finite element algorithm that handles the specific nonlinear coupling and geometric constraints of ferronematics, offering a practical tool for simulating these materials in optics, sensors, and telecommunications applications.