Table of Contents
Fetching ...

Two and three dimensional $H^2$-conforming finite element approximations without $C^1$-elements

Mark Ainsworth, Charles Parker

TL;DR

This paper tackles the challenge of obtaining $H^2$-conforming finite element approximations in 2D and 3D without $C^1$-elements by introducing a novel mixed variational formulation that uses only $H^1$-regular spaces. By introducing the auxiliary variable $\boldsymbol{\gamma}=\nabla w$ and enforcing $\boldsymbol{\gamma}=\nabla w$ through a Lagrange multiplier in $(\mathrm{Im}\boldsymbol{\Xi})^*=\boldsymbol{H}_{\Gamma}(\mathrm{curl};\Omega)^*$, the authors recover $H^2$-conforming discretizations using standard $H^1$-spaces and curl-conforming spaces. The discrete scheme employs choices of $\tilde{\mathbb{W}}_{\Gamma}$, $\boldsymbol{\mathbb{G}}_{\Gamma}$, and $\boldsymbol{\mathbb{Q}}_{\Gamma}$, together with an iterated penalty method to avoid constructing a basis for $\mathrm{Im}\boldsymbol{\Xi}_X$, and it yields $H^2_{\Gamma}(\Omega)$ solutions that coincide with the original $H^2$-conforming approximations. The approach is demonstrated through 2D and 3D examples, including Kirchhoff plates (static and dynamic) and $H^2$-projections, and extended to time-dependent problems with careful treatment of stability constants. The work also establishes uniform stability results for the Morgan–Scott space and provides right-inverse constructions for curl and the $\Xi_X$ operator, linking to previous work and generalizing it to broader settings.

Abstract

We develop a method to compute $H^2$-conforming finite element approximations in both two and three space dimensions using readily available finite element spaces. This is accomplished by deriving a novel, equivalent mixed variational formulation involving spaces with at most $H^1$-smoothness, so that conforming discretizations require at most $C^0$-continuity. The method is demonstrated on arbitrary order $C^1$-splines.

Two and three dimensional $H^2$-conforming finite element approximations without $C^1$-elements

TL;DR

This paper tackles the challenge of obtaining -conforming finite element approximations in 2D and 3D without -elements by introducing a novel mixed variational formulation that uses only -regular spaces. By introducing the auxiliary variable and enforcing through a Lagrange multiplier in , the authors recover -conforming discretizations using standard -spaces and curl-conforming spaces. The discrete scheme employs choices of , , and , together with an iterated penalty method to avoid constructing a basis for , and it yields solutions that coincide with the original -conforming approximations. The approach is demonstrated through 2D and 3D examples, including Kirchhoff plates (static and dynamic) and -projections, and extended to time-dependent problems with careful treatment of stability constants. The work also establishes uniform stability results for the Morgan–Scott space and provides right-inverse constructions for curl and the operator, linking to previous work and generalizing it to broader settings.

Abstract

We develop a method to compute -conforming finite element approximations in both two and three space dimensions using readily available finite element spaces. This is accomplished by deriving a novel, equivalent mixed variational formulation involving spaces with at most -smoothness, so that conforming discretizations require at most -continuity. The method is demonstrated on arbitrary order -splines.
Paper Structure (29 sections, 25 theorems, 146 equations, 6 figures, 1 table)

This paper contains 29 sections, 25 theorems, 146 equations, 6 figures, 1 table.

Key Result

Lemma 3.1

For every $\boldsymbol{\theta} \in \boldsymbol{H} _{\Gamma}(\mathrm{curl};\Omega)$, there exists $(\tilde{v}, \boldsymbol{\psi} ) \in H^1_{\Gamma}(\Omega) \times \boldsymbol{G} _{\Gamma}(\Omega)$ such that where $C_1$ and $C_2$ are positive constants independent of $\boldsymbol{\theta}$. Consequently, $\mathop{\mathrm{Im}}\nolimits \boldsymbol{\Xi} = \boldsymbol{H} _{\Gamma}(\mathrm{curl};

Figures (6)

  • Figure 1: (a) The displacement $\tilde{w}_X$ and (b) $|\mathop{\mathrm{\mathbf{grad}}}\nolimits \tilde{w}_X|$ for the $p=10$ solution to the Kirchhoff plate problem \ref{['eq:kirchhoff-ex-pointload']}.
  • Figure 2: Relative $H^2$ errors and for the projection problem \ref{['eq:3d-projection-problem']} with degree-$p$$C^1$-spline elements \ref{['eq:w-c1-spline']} (a) on a sequence of meshes with $2 \leq p \leq 9$ and (b) on the single mesh $\mathcal{T}_2$ with $2 \leq p \leq 12$. The number next to each dashed line in (a) indicates its slope.
  • Figure 3: (a) The initial acceleration $\tilde{w}_0^{[2]}$\ref{['eq:newmark-initial-dtt']} and (b) $|\mathop{\mathrm{\mathbf{grad}}}\nolimits \tilde{w}_0^{[2]} |$ for the $p=10$ solution to the dynamic Kirchhoff plate problem.
  • Figure 4: (a) The number of iterated penalty (IP) iterations to compute the initial acceleration $w_0^{[2]}$ given by \ref{['eq:newmark-initial-dtt']} and the max number of iterations to advance the acceleration $w_n^{[2]}$ via \ref{['eq:newmark-dtt-update']}; and (b) the maximum relative energy deviation \ref{['eq:newmark-energy-deviation']} computed using $\mathop{\mathrm{\mathbf{grad}}}\nolimits \tilde{w}_n$ and $\boldsymbol{\gamma} _n$. The metrics in (a) and (b) are computed over 251 time steps.
  • Figure 5: The $p=10$ displacement $\tilde{w}_n$ and its gradient (a-b) and velocity $\tilde{w}_n^{[1]}$ and its gradient (c-d) for the dynamic Kirchhoff plate problem at time $t=0.0502$.
  • ...and 1 more figures

Theorems & Definitions (46)

  • Lemma 3.1
  • proof
  • Corollary 3.1
  • proof
  • Lemma 3.2
  • proof
  • Theorem 3.1
  • proof
  • Corollary 3.2
  • proof
  • ...and 36 more