🔢 mathematics

Fully discrete error analysis of finite element discretizations of time-dependent Stokes equations in a stream-function formulation

This paper establishes best approximation error estimates for fully discrete Galerkin solutions of the time-dependent Stokes equations in a stream-function formulation using discontinuous Galerkin time discretization and a general space discretization framework, applicable to methods like $C^1$ and $C^0$ interior penalty schemes without requiring additional regularity assumptions.

Original authors: Dmitriy Leykekhman, Boris Vexler, Jakob Wagner

Published 2026-05-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Dmitriy Leykekhman, Boris Vexler, Jakob Wagner

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 you are trying to predict how a thick, slow-moving fluid (like honey or oil) flows through a container. In the world of physics and engineering, this is described by something called the Stokes equations. These equations are like a complex recipe that tells us how the fluid moves (velocity) and how much pressure is pushing it around.

Usually, solving this recipe on a computer is tricky. It's like trying to bake a cake where the instructions for the "flour" (velocity) and the "sugar" (pressure) are so tightly mixed that if you make a tiny mistake in measuring the sugar, your whole cake (the velocity) turns out ruined. Standard computer methods often struggle with this "pressure sensitivity," leading to inaccurate results if the data isn't perfect.

The Paper's Big Idea: The "Shadow" Method
Instead of trying to calculate the fluid's movement and pressure separately, the authors of this paper use a clever trick called the stream-function formulation.

Think of the fluid's movement not as a messy pile of arrows, but as a single, smooth "shadow" or "contour map" (called $\psi$ ).

In this method, the fluid's speed is derived directly from the shape of this shadow.
Because the shadow is built to be smooth, the fluid automatically obeys the rule that it can't be created or destroyed (it's "divergence-free").
This approach completely removes the "sugar" (pressure) from the main equation. The computer only has to solve for the shadow, and the fluid flow falls out naturally.

The Challenge: The "Time-Traveling" Shadow
The paper focuses on fluids that change over time (time-dependent). This adds a layer of complexity: the shadow isn't just a static picture; it's a movie. The authors are trying to figure out how to chop this movie into tiny frames (time steps) and tiny pixels (space steps) to simulate it on a computer without the picture getting blurry or distorted.

They use a specific technique called Discontinuous Galerkin (dG) for the time steps. Imagine watching a movie where, instead of smooth motion, you see a series of slightly jerky snapshots. This method allows the "shadow" to jump a little bit between frames, which gives the computer more flexibility and stability, preventing the simulation from crashing.

What They Proved: The "Best Guess" Guarantee
The core of the paper is a mathematical proof that answers a very practical question: "If I use this specific way of chopping up the movie and the pixels, how close will my computer simulation be to the real, perfect physics?"

They established a "Best-Approximation" guarantee. Here is the analogy:
Imagine you are trying to draw a perfect circle on a grid of graph paper.

The Error: The difference between your drawing and a perfect circle.
The Guarantee: The authors proved that the error in their computer simulation is no worse than the error you would get if you just took the best possible drawing you could make with that specific grid and those specific time steps.

In other words, their method is as good as it possibly can be given the tools (the grid size and time steps) you are using. You aren't losing accuracy because of the method itself; you are only losing accuracy because of the resolution you chose.

Key Findings in Plain English:

No Extra Assumptions Needed: Many math papers require the fluid to be "extra smooth" or the container to be perfectly round to get good results. This paper says, "Nope, our method works even if the data is messy or the container is a weird polygon." It works with the "natural" level of smoothness you'd expect in real life.
Pressure-Proof: Because they removed the pressure from the main equation, their method is "pressure-robust." If you add a fake, invisible force that looks like pressure but doesn't actually push the fluid, their method ignores it completely. Standard methods would get confused and produce errors.
Flexible Tools: Their proof works for a wide variety of ways to build the grid (the "pixels"), not just one specific type. This makes their result useful for many different engineering software packages.

The "Gotcha" (The Counterexample)
In the appendix, the authors play devil's advocate. They show a specific, tricky mathematical case where the "shadow" changes so violently that you cannot expect the computer to predict its speed perfectly, no matter how good your grid is. This is important because it tells us the limits of the theory: we can't demand more accuracy than the physics allows.

The Takeaway
This paper provides a solid, mathematical safety net for engineers and scientists who want to simulate time-changing fluid flows using the "stream-function" trick. It proves that if you use their specific time-stepping method, your computer results will be as accurate as the grid and time-steps you choose allow, without being ruined by messy pressure calculations or weird boundary shapes. It's a "best possible outcome" guarantee for a specific way of solving fluid problems.

Problem Statement
This paper addresses the fully discrete error analysis of finite element discretizations for the time-dependent Stokes equations in two dimensions, specifically utilizing a stream-function formulation. The classical Stokes problem involves finding velocity $\mathbf{u}$ and pressure $p$ satisfying the momentum and continuity equations. In the stream-function approach, the velocity is represented as the vectorial curl of a scalar stream-function $\psi$ ( $\mathbf{u} = \text{curl}(\psi)$ ), which automatically satisfies the divergence-free condition. This transformation reduces the system to a single fourth-order parabolic equation involving the spatial biharmonic operator ( $\Delta^2$ ) and the time derivative of the Laplacian ( $\partial_t \Delta$ ).

The authors highlight a critical limitation in standard finite element methods (such as Taylor-Hood or MINI elements) for Stokes problems: the error estimates for velocity often depend on the pressure approximation. Consequently, perturbations in the source term by a gradient field may not be absorbed by the discrete pressure, leading to suboptimal velocity errors. The stream-function formulation offers a pathway to "pressure-robust" methods, but a rigorous theory for fully discrete error estimates (combining space and time discretization) for this specific fourth-order time-dependent problem has been lacking.

Methodology
The paper employs a fully discrete Galerkin framework with the following components:

Time Discretization: The authors utilize the Discontinuous Galerkin (dG) method of arbitrary degree $r$ . This involves partitioning the time interval into subintervals and using piecewise polynomials, allowing for jumps at the time nodes.
Space Discretization: The spatial discretization is presented in a general framework rather than being tied to a specific element type. The analysis relies on a symmetric, elliptic discrete bilinear form $a_h(\cdot, \cdot)$ $a_{h} (\cdot, \cdot)$ and a discrete operator $A_h$ $A_{h}$ . The framework is designed to accommodate various methods, provided they satisfy specific Galerkin orthogonality conditions. The paper explicitly covers:
- Conformal $C^1$ finite elements (e.g., Argyris elements).
- $C^0$ interior penalty methods (using standard Lagrange elements of degree $\ge 2$ ).
Regularity Analysis: A significant portion of the work is dedicated to establishing the regularity of the continuous solution $\psi$ . The authors prove existence, uniqueness, and stability results for the continuous problem under minimal regularity assumptions on the data ( $f \in L^2(I; H^{-1}(\Omega))$ and $\psi_0 \in H^2_0(\Omega)$ ). Crucially, they demonstrate that while the solution possesses certain stability properties, higher regularity of the time derivative $\partial_t \psi$ (specifically in $L^2$ or $H^{-1}$ ) cannot be guaranteed under these minimal assumptions, a fact proven via a counterexample in Appendix A.

Key Contributions and Results
The primary contribution is the establishment of best-approximation-type error estimates for the fully discrete solution $\psi_{kh}$ . The main result (Theorem 4.5) states that the error in the gradient of the stream-function is bounded by the sum of three approximation errors:
$\|\nabla(\psi - \psi_{kh})\|_{L^2(I \times \Omega)} \le C \left( \|\nabla(\psi - \chi_{kh})\|_{L^2(I \times \Omega)} + \|\nabla(\psi - R_h\psi)\|_{L^2(I \times \Omega)} + \|\nabla(\psi - \pi_k\psi)\|_{L^2(I \times \Omega)} \right)$
where:

$\chi_{kh}$ is an arbitrary element from the discrete space.
$R_h$ is the Galerkin projection associated with the stationary biharmonic problem.
$\pi_k$ is a specific time projection.

By applying standard approximation properties of these projections under the established regularity assumptions, the authors derive optimal convergence rates on convex polygonal domains:
$\|\nabla(\psi - \psi_{kh})\|_{L^2(I \times \Omega)} \le C(k + h^2) \left( \|f\|_{L^2(I; H^{-1}(\Omega))} + \|\psi_0\|_{H^2_0(\Omega)} \right)$
This implies an $O(k)$ convergence in time and $O(h^2)$ in space for the velocity error (since $\|\mathbf{u} - \mathbf{u}_{kh}\| = \|\nabla(\psi - \psi_{kh})\|$ ).

Significance and Claims
The paper positions itself as a foundational step in the numerical analysis of stream-function formulations for time-dependent problems.

Pressure Robustness: The derived estimates do not contain pressure terms, addressing the known issue where standard methods suffer from velocity errors dependent on pressure approximations. The numerical experiments confirm that the stream-function method remains robust even when the source term is modified by a gradient field, whereas standard mixed methods (MINI elements) show degraded performance.
Minimal Regularity Assumptions: The analysis does not require additional regularity assumptions beyond those naturally provided by the domain and data. The authors explicitly argue that previous works often imposed overly strict assumptions (e.g., requiring $\partial_{tt}\psi$ to be bounded) to achieve similar estimates.
Generality: The framework is applicable to a wide variety of spatial discretization methods, provided they satisfy the necessary orthogonality conditions, making the results versatile for conformal $C^1$ and non-conformal $C^0$ interior penalty methods.
Optimal Control: The authors note that the low regularity assumptions make these results suitable for application in optimal control problems.

The paper concludes with numerical experiments that illustrate the theoretical convergence rates and demonstrate the pressure-robustness of the proposed method compared to standard mixed finite element formulations. The authors acknowledge this as an initial step, with future work planned to extend these results to surface flows, multiply connected domains, and the Navier-Stokes equations.

More like this