Linearly Implicit Parametric FE Scheme

• The paper demonstrates that linearization and extrapolation techniques enable solving complex nonlinear PDEs via a single linear finite-element system at each time step.
• The scheme employs explicit parametric discretization to accurately model evolving geometries in applications like surface flows and ALE-based multiphysics.
• The approach guarantees unconditional stability, energy dissipation, and optimal convergence using advanced time-stepping coupled with mesh regularization.

A linearly implicit parametric finite-element scheme refers to a broad class of time-stepping and spatial discretization methods for evolutionary PDEs—especially geometric flows and moving-domain problems—in which nonlinear (or geometry-dependent) terms are treated using extrapolation or linearization, such that, at each time step, only a linear finite-element system must be solved. Parametric refers to either an explicit surface or curve parametrization (for interface/moving boundary problems), or to a referenced field parameter (as in reference-state linearizations for fluid dynamics). These schemes deliver significant computational advantages in geometric evolution equations, multiphysics moving-interface PDEs, and incompressible/compressible flow regimes, maintaining stability and high-order accuracy in both space and time.

1. Fundamental Framework and Formulation

Linearly implicit parametric finite-element schemes are distinguished by two core features:

• Parametric discretization: The evolving domain (curve/surface or mapped reference region) is represented via an explicit parametrization (e.g., $X(q,t)$ for surfaces, $X(\rho, t)$ for curves) or a mapped ALE coordinate system. Finite-element spaces are then defined on these parametric representations, enabling mesh alignment with evolving geometry (Kovács et al., 2017, Jiang et al., 2023, Bao et al., 26 Jun 2025, Schwarzacher et al., 2023).
• Linearly implicit time discretization: At each time step, the discretization is designed so the only required algebraic solve is linear. This is achieved through either extrapolation in nonlinear terms (e.g., explicit geometry for assembling matrices, or explicit convection), Jacobian–linearization in the direction of a reference parameter field, or semi-implicit partitioning of the fluxes (Qi, 2018, Kučera et al., 2020).

The general workflow is as follows:

1. Derive the weak variational form on the parametric or mapped domain.
2. Select finite-element spaces consistent with the geometry and regularity of the evolving field(s).
3. Time-discretize evolution equations using linearly implicit schemes (e.g., $k$-step BDF, Crank–Nicolson, backward Euler, leapfrog-CN).
4. Explicitly evaluate (extrapolate/freeze) all nonlinear and geometry-dependent coefficients at previous time levels or reference states, ensuring a linear system at each time step.

2. Prototypical Schemes and Applications

Linearly implicit parametric finite-element schemes are realized in several major settings:

• Geometric and curvature-driven surface/curve flows: For mean curvature flow, surface diffusion, Willmore flow, and related geometric gradient flows, parametric finite-element (PFEM) and evolving-surface FEM (ESFEM) employ linear spatial discretization on triangulated evolving domains. The temporal evolution is handled via linearly implicit BDF, Crank–Nicolson, or equivalently structured energy-stable schemes, leading to per-step linear algebraic systems incorporating geometry at frozen or extrapolated levels (Bao et al., 26 Jun 2025, Jiang et al., 2023, Kovács et al., 2017).
• Moving boundary, ALE-based multiphysics: In fluid–structure interaction or time-dependent domains, an ALE mapping from a reference domain is exploited; the governing equations are weakly formulated on the reference mesh. Linearly implicit backward Euler (in which geometric coefficients are frozen at the previous step) achieves unconditional energy dissipation and optimal first-order accuracy (Schwarzacher et al., 2023).
• Compressible/incompressible flows and stiff hyperbolic limits: Reference-state-based linearization, where each time step uses the Jacobian of the flux at a reference state $U^*$, leads to a linearly implicit FE method maintaining asymptotic-preserving (AP) properties in the low-Mach or singular perturbation limit (Kučera et al., 2020).
• Second-order parabolic PDEs and stochastic evolution: Weak Galerkin and Rothe–Euler schemes, including double-valued edge functions and stabilization terms, are cast into linearly implicit FE time-stepping (e.g., $\theta$-schemes), retaining optimal convergence and stability (Qi, 2018, Cioica et al., 2015).

A representative table illustrates this diversity:

Application Class	Discretization Approach	Time-Stepping Scheme
Geometric/mean curvature flow	PFEM/ESFEM on evolving parametric meshes	Linearly implicit BDF, CN
Fluid–structure–ALE	FE on mapped reference domain (ALE)	Linearly implicit Euler
Weakly compressible Euler	FE with reference-state Jacobian splitting	Linearly implicit BE (AP)
SPDEs (Rothe–Euler)	Hilbert/FE + residual-based linear solves	Linearly implicit Euler

3. Algorithmic Structure and Matrix Formulations

Per time step, the discrete system typically takes the form

$$\left[M + \alpha A(\mathcal{G}^{old})\right] U^{n+1} = \text{rhs}$$

where $M$ is the (possibly lumped) mass matrix, $A(\mathcal{G}^{old})$ is a stiffness or geometric operator assembled from geometry at extrapolated or frozen time-level $\mathcal{G}^{old}$, and "rhs" collects all explicit, previous-step, or reference-dependent contributions.

• Geometry freezing/extrapolation: Geometry- or nonlinearity-dependent matrices (e.g., mass, stiffness, curvature terms) are assembled at either a fixed earlier mesh, an extrapolated mesh (using previous several time steps for BDF-$k$), or a reference parameter field (Kovács et al., 2017, Kučera et al., 2020, Jiang et al., 2023).
• Linear system solution: One global linear algebraic system is solved per time step, of order proportional to the total number of spatial degrees of freedom. Saddle-point structures, block partitioned by field (e.g., position, curvature, velocity) are common (Bao et al., 26 Jun 2025, Qi, 2018).
• Stabilization and regularization: Penalty terms, mass-lumping, or tangential-velocity control are included to ensure discrete stability, optimal mesh regularity, and unconditional energy decay (Bao et al., 26 Jun 2025, Jiang et al., 2023).

4. Stability, Energy Dissipation, and Error Analysis

A distinguishing feature of linearly implicit parametric FE schemes is the ability to ensure unconditional (A-)stability and, in many cases, monotonic discrete energy dissipation independent of time step or mesh size:

• Unconditional stability: For $\theta$-schemes with $1/2\le\theta\le1$ (backward Euler, CN), or linearly implicit BDF, stability is unconditional in the energy norm (Qi, 2018, Kovács et al., 2017).
• Energy dissipation: Discrete monotonicity of geometric energy functionals (e.g., Willmore energy, total perimeter) can be proven for schemes designed using variational splitting and mass-lumping (Bao et al., 26 Jun 2025, Jiang et al., 2023).
• Error estimates: Sharp bounds are typically of the form

$\|u^{n} - u_{h}^{n}\| \leq C(h^{k} + \tau^{p})$

with $k$ dictated by FE degree and $p$ by temporal order, under appropriate regularity and consistency assumptions (Qi, 2018, Kovács et al., 2017, Schwarzacher et al., 2023). For stochastic or inexact-elliptic sub-solvers, per-time-step tolerances must be calibrated to preserve global order (Cioica et al., 2015).

In the context of asymptotic-preserving (AP) schemes for stiff flow limits, additional estimates guarantee correct low-Mach or singular limit behavior and second-order accuracy for key fields (Kučera et al., 2020).

5. Tangential Velocity, Mesh Quality, and Regularization

Mesh regularity is critical in parametric and ALE methods due to their direct dependence on the nodal geometry:

• Tangential velocity control: Auxiliary constraints or penalty formulations are introduced to suppress tangential motion and prevent mesh degeneration. This includes tangential fields and penalty terms projected onto local tangent frames, with adaptively updated penalty parameters (Bao et al., 26 Jun 2025).
• Mesh regularization strategies: Two main strategies emerge:
  • Occasional re-equilibration via first-order or trivial-flow steps when mesh quality degrades beyond a threshold.
  • Continuous equidistribution enforced as an interlaced step, e.g., trivial-flow PFEM stage enforcing equal edge lengths (Jiang et al., 2023).
• Energy stability under mesh regularization: Strategies can be designed to interlace mesh regularization with time steps without sacrificing discrete energy dissipation, validated in unconditional interlaced energy-stability results (Jiang et al., 2023).

6. Extensions, Representative Results, and Numerical Validation

Comprehensive analyses and demonstrations across evolving geometries, multiphysics, and hyperbolic limits (weakly compressible, low–Froude) are presented:

• Geometric flows: Linearly implicit ESFEM and PFEM enable high-fidelity simulation of mean curvature, surface diffusion, and Willmore flows with unconditionally stable and energy-dissipating discrete trajectories. High-order accuracy is observed until spatial errors dominate at fine mesh scales (Kovács et al., 2017, Jiang et al., 2023, Bao et al., 26 Jun 2025).
• Fluid–structure ALE: Semi-implicit linearly implicit FE-ALE solvers demonstrate linear convergence in both $h$ and $\tau$, energy dissipation, geometric conservation, and computational speedup versus fully nonlinear implicit methods (Schwarzacher et al., 2023).
• Weakly compressible flows: AP linearly implicit methods attain uniform convergence in the low-Mach limit and recover correct divergence-free dynamics as singular parameters vanish (Kučera et al., 2020).
• Stochastic PDEs: Linearly implicit FE–Rothe schemes with residual-based inexact solvers yield optimal convergence provided tolerances scale with time-step, as established via a discrete Grönwall approach (Cioica et al., 2015).
• Error metrics: For parametric schemes on curves and surfaces, shape metrics such as manifold distance or Hausdorff distance yield superior asymptotic convergence properties for BGN-based second-order schemes (Jiang et al., 2023).

7. Impact and Theoretical Significance

The linearly implicit parametric finite-element paradigm delivers robust, high-order, and computationally efficient algorithms for evolutionary geometric PDEs, interface problems, and stiff singular limits. Its major impacts include:

• Enabling practical large-scale simulations of evolving interfaces and coupled multiphysics without sacrificing stability or accuracy.
• Bridging the gap between efficient time stepping (via linear systems only) and sophisticated geometric or multiphysics couplings which typically introduce severe nonlinearities.
• Underpinning a variety of energy-stable and AP schemes critical for application areas ranging from materials science (e.g., Willmore/MCF), biophysics (membrane models), to compressible/incompressible transition flows.

Detailed algorithmic and theoretical analyses—spanning stability, convergence, and mesh quality properties—are rigorously established in the cited literature (Qi, 2018, Kovács et al., 2017, Bao et al., 26 Jun 2025, Jiang et al., 2023, Schwarzacher et al., 2023, Kučera et al., 2020, Cioica et al., 2015).