🔢 mathematics

A Hybridized Staggered Discontinuous Galerkin--Mixed Finite Element Method for Strain Gradient Elasticity

This paper proposes a hybridized staggered discontinuous Galerkin--mixed finite element method for strain gradient elasticity that achieves parameter-robust stability and optimal convergence by recasting the fourth-order problem into a first-order system with strong symmetry enforcement and a hybridized formulation that minimizes global coupling.

Original authors: Bohan Yang, Eric T. Chung

Published 2026-09-15

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Bohan Yang, Eric T. Chung

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where the rules of how materials bend and stretch change depending on how small you look at them. For over a century, engineers have relied on classical elasticity, a set of laws that treats materials like a continuous, seamless fabric. These laws work perfectly for bridges, buildings, and large machines. However, they fail when applied to the microscopic structures found in modern technology, such as the tiny wires in microchips or the thin beams in medical sensors. At these scales, materials exhibit "size effects," behaving differently simply because they are small. A thin wire might twist much harder than classical physics predicts, or a microbeam might resist bending in unexpected ways. To explain this, scientists developed strain gradient elasticity, a more advanced theory that adds a specific "length scale" to the equations. This extra parameter accounts for the internal structure of the material, allowing the math to capture those mysterious size-dependent behaviors.

The challenge, however, is that these advanced equations are incredibly difficult to solve on a computer. They are fourth-order problems, meaning they involve complex relationships between how a material deforms and how that deformation changes across space. Traditional methods for solving them often stumble when the material is nearly incompressible (like rubber) or when the length scale is very small, leading to inaccurate results or "numerical boundary layers" where the computer creates artificial errors near the edges of the object. Researchers have tried various approaches, but many require delicate tuning of parameters or fail to provide a direct way to calculate the total stress, a crucial quantity for engineers designing safe micro-devices.

In a new study, researchers Bohan Yang and Eric T. Chung have proposed a fresh way to tackle this problem. They developed a hybridized method that combines two powerful mathematical techniques: the staggered discontinuous Galerkin method and the mixed finite element method. Instead of trying to solve for the displacement of the material alone, their approach introduces several new variables to represent the different types of stress within the material. They treat the total stress, along with scaled versions of the internal forces, as independent unknowns. By doing this, they recast the difficult fourth-order problem into a system of simpler, first-order equations. This shift allows them to use specific types of mathematical spaces that naturally handle the complexity of the problem without needing artificial fixes or penalty terms.

The core of their innovation lies in how they organize the calculation. They divide the material's geometry into a grid of triangles and then create a second, overlapping grid within it. On this dual structure, they enforce different rules for continuity, allowing the mathematical pieces to fit together in a way that preserves the physical symmetry of the stress. This design ensures that the method remains stable and accurate regardless of how small the material's length scale is or how close the material gets to being incompressible. A key feature of their work is that they can mathematically prove the method is robust; the accuracy does not degrade when the parameters change, a common failure point for other methods. Furthermore, they introduced a "hybridized" version of their scheme that allows them to solve for most variables locally within each small triangle, leaving only a few global variables to be solved together. This dramatically reduces the computational cost, making the method efficient enough for practical use.

The researchers tested their method with numerical experiments, simulating scenarios with both smooth solutions and those featuring strong boundary layers, which are sharp changes in behavior near the edges of the material. In every case, their method achieved the predicted rates of accuracy and remained stable across a wide range of material properties, including extreme values for stiffness and incompressibility. Unlike previous displacement-based methods that struggle when the length scale is tiny, their approach maintained optimal accuracy without creating artificial errors. They also demonstrated that their method provides a direct and accurate calculation of the total stress, a result that is often difficult to obtain with other techniques. While the current work focuses on two-dimensional shapes with specific boundary conditions, the success of this approach suggests a promising path forward for modeling complex micro-scale materials. The study confirms that by rethinking how these equations are structured and solved, it is possible to overcome the long-standing difficulties of strain gradient elasticity, offering a reliable tool for the next generation of micro-engineering.

Problem Statement
The paper addresses the numerical approximation of the strain gradient elasticity (SGE) model, specifically the formulation proposed by Aifantis and collaborators. This model is a fourth-order singularly perturbed problem governed by a material length scale parameter $\iota \in (0, 1]$ and the Lamé constants $\lambda$ and $\mu$ . The governing equation, $-(1 - \iota^2\Delta)(\mu\Delta u + (\lambda + \mu)\nabla\text{div} u) = f$ , presents two primary challenges for discretization:

High-order continuity: As a fourth-order problem, standard finite element methods require $H^2$ -conformity. While $C^1$ -conforming elements exist, they are complex to construct and computationally expensive. Nonconforming methods often struggle with stability or require careful tuning of penalty parameters.
Parameter robustness: The problem is singularly perturbed with respect to $\iota$ . As $\iota \to 0$ , the model reduces to classical elasticity, but the higher-order Dirichlet boundary condition ( $\partial_n u = 0$ ) cannot be satisfied by the classical solution, leading to a physical boundary layer of thickness $O(\iota)$ . Displacement-based methods that enforce this condition strongly on the discrete space often suffer from a "numerical boundary layer" of thickness governed by the mesh size $h$ , degrading convergence rates to $O(h^{1/2})$ when $\iota \ll 1$ . Additionally, as $\lambda \to \infty$ (the incompressible limit), methods may suffer from volumetric locking.

Methodology
The authors propose a hybridized staggered discontinuous Galerkin (SDG)–mixed finite element method. The approach involves the following steps:

First-Order Reformulation: The fourth-order equation is recast as a first-order system by introducing the total stress $\sigma$ and two auxiliary variables: the scaled Cauchy stress $\tilde{\sigma}$ and the scaled hyper stress $\chi$ . This results in a four-field system involving displacement $u$ , total stress $\sigma$ , scaled Cauchy stress $\tilde{\sigma}$ , and scaled hyper stress $\chi$ . Symmetry is imposed strongly on all three stress tensors.
Discrete Spaces:
- The displacement $u$ and total stress $\sigma$ are discretized using Staggered Discontinuous Galerkin (SDG) spaces. These spaces utilize an Alfeld split of the primal mesh to create a dual mesh, enforcing continuity across primal edges for $u$ and normal continuity across dual edges for the total stress $\sigma$ .
- The scaled stresses $\tilde{\sigma}$ and $\chi$ are discretized using Raviart–Thomas (RT) pairs and discontinuous Galerkin spaces, respectively, with symmetry imposed.
Hybridization: To reduce the global system size, the inter-element normal continuity of the total stress and scaled hyper stress is relaxed and reimposed weakly via Lagrange multipliers defined on the mesh skeleton (the dual edges). This allows for local static condensation, where the internal degrees of freedom are eliminated locally, leaving only the multipliers as globally coupled unknowns.
Boundary Conditions: The higher-order Dirichlet boundary condition is incorporated naturally into the variational formulation rather than being imposed strongly on the discrete space, thereby avoiding the numerical boundary layer.

Key Contributions and Theoretical Results
The paper provides a rigorous theoretical analysis supporting the proposed method:

Stability: The authors establish an inf-sup condition on the symmetric subspace and a discrete Korn inequality. These tools are used to prove the algebraic stability of the scheme.
Convergence:
- Under sufficient regularity, the method achieves optimal convergence rates for the displacement and stresses.
- Crucially, the error estimates are uniform in both $\iota$ and $\lambda$ . The method retains optimal convergence rates even when $\iota \ll 1$ , avoiding the degradation seen in displacement-based methods.
- The analysis explicitly tracks the dependence on the regularity of the solution and the parameters.
Hybridization: The paper develops the hybridized scheme, proving that the two multipliers (approximating the traces of $u$ and $\iota\tilde{\sigma}$ ) satisfy their own inf-sup conditions, ensuring the stability of the reduced global system.
Incompressible Limit: A modification is provided for the case $\lambda \to \infty$ (incompressible limit), involving a decomposition of the stress to handle the global normalization constraint required for well-posedness.

Numerical Results
Two numerical experiments are reported to validate the theory:

Smooth Solution: A test case with a smooth, divergence-free solution confirms optimal convergence rates in both $L^2$ and discrete $H^1$ norms for the displacement, and optimal rates for the total stress. The results demonstrate robustness across a wide range of $\lambda$ (from 1 to $10^8$ ) and $\iota$ (from 1 to $10^{-8}$ ).
Boundary Layer: A test case with a strong boundary layer (where the exact solution of the SGE problem differs from the classical solution near the boundary) confirms that the method maintains optimal convergence rates and parameter robustness even when $\iota$ is very small ( $10^{-6}, 10^{-8}$ ).

Significance
The paper claims to present the first four-field mixed scheme for strain gradient elasticity supported by a rigorous analysis. Its significance lies in:

Avoiding the Numerical Boundary Layer: By incorporating the higher-order boundary condition variationally rather than strongly, the method retains optimal convergence rates for small $\iota$ , a known difficulty for displacement-based methods.
Parameter Robustness: The scheme is stable and convergent uniformly in both the length scale parameter $\iota$ and the Lamé constant $\lambda$ , addressing issues of singular perturbation and incompressibility simultaneously.
Direct Stress Approximation: Unlike displacement-based methods, this mixed formulation provides a direct approximation of the total stress, a quantity of practical interest, without requiring post-processing of higher-order derivatives.
Computational Efficiency: The hybridized version reduces the global system to only two multipliers on the mesh skeleton, significantly lowering the computational cost compared to the non-hybridized four-field system.

The work is currently restricted to two dimensions and homogeneous doubly clamped boundary conditions, with extensions to three dimensions and general boundary conditions noted as future work.

More like this