🔢 mathematics

Quadrature error estimates on non-matching grids in a fictitious domain framework for fluid-structure interaction problems

This paper establishes the well-posedness of a fictitious domain formulation for fluid-structure interaction problems using distributed Lagrange multipliers and provides quadrature error estimates for the coupling term when computed on non-matching grids via approximate integration.

Original authors: Daniele Boffi, Fabio Credali, Lucia Gastaldi

Published 2026-06-30

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Daniele Boffi, Fabio Credali, Lucia Gastaldi

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 simulate how a flexible rubber ball moves through a pool of water. In the world of computer simulations, this is called Fluid-Structure Interaction (FSI).

To do this, computers usually need to break the world into tiny puzzle pieces (a "mesh").

The Water: Needs a grid of puzzle pieces to calculate how the water flows.
The Ball: Needs its own grid of puzzle pieces to calculate how it stretches and bends.

The Problem: Two Different Jigsaw Puzzles

In many real-world scenarios, the ball moves and changes shape. If you try to force the water's puzzle pieces to perfectly match the ball's puzzle pieces every time the ball moves, the pieces get twisted, stretched, and distorted until the computer simulation crashes.

To avoid this, scientists use a clever trick called the Fictitious Domain Method.

Imagine the water has a giant, rigid grid that never changes shape.
The ball is a separate, smaller grid that floats on top of the water grid.
The ball doesn't care about the water's grid lines; it just floats wherever it wants.

The Catch: To make the physics work, the computer must tell the water, "Hey, where the ball is, you need to move exactly like the ball." This requires a "handshake" or a coupling between the two different grids.

The Challenge: The "Approximate" Handshake

To calculate this handshake, the computer has to multiply numbers from the ball's grid by numbers from the water's grid.

The Perfect Way: The computer could cut the water grid into tiny pieces wherever the ball touches it, creating a super-fine, custom map for the calculation. This is mathematically perfect but very slow and computationally expensive (like trying to cut a pizza into a million tiny slices just to feed one person).
The Approximate Way: The computer just uses the original, coarse grids and guesses the answer by sampling a few points. This is much faster, but it introduces a small amount of error (a "quadrature error").

The Big Question: If we use the fast, approximate way, does the simulation break? Is the answer still trustworthy?

What This Paper Does

The authors of this paper, Daniele Boffi, Fabio Credali, and Lucia Gastaldi, decided to investigate this "Approximate Way." They didn't just guess; they did the heavy mathematical lifting to prove two main things:

It's Stable: They proved mathematically that even if you use the fast, approximate method (without cutting the grids), the simulation remains stable. The numbers won't explode, and a unique, valid solution still exists.
It's Accurate (with conditions): They calculated exactly how much error is introduced by this approximation.
- Scenario A (Simple Coupling): If the handshake is simple (just matching positions), the error is small and manageable, provided the ball's grid is fine enough.
- Scenario B (Complex Coupling): If the handshake is complex (involving how the ball is stretching, not just where it is), the error can be larger. To keep the answer accurate, the ball's grid must not only be fine, but it must also be much finer than the water's grid.

The Analogy: The Translator

Think of the water grid and the ball grid as two people speaking different languages.

The Exact Method: They hire a translator who sits between them, listens to every single word, and translates it perfectly. This is accurate but takes a long time.
The Approximate Method: They hire a translator who only listens to every 10th word and guesses the rest.
- The paper proves that if the conversation is simple (just "I am here"), the guesser does a great job.
- If the conversation is complex (discussing "how I am stretching"), the guesser might make mistakes unless the person speaking the ball's language speaks very slowly and clearly (a very fine grid).

The Results

The authors ran computer tests to confirm their math.

They found that for simple interactions, the "fast" method works just as well as the "perfect" method.
For complex interactions, the "fast" method works well only if the ball's grid is significantly denser than the water's grid. If the grids are the same size, the fast method loses accuracy.

The Bottom Line

This paper gives scientists the green light to use faster, cheaper simulation methods for fluid-structure problems. It tells them: "You can skip the expensive step of cutting the grids together, but you need to be careful about how fine your grids are, depending on how complex the interaction is."

They validated this with numerical tests, showing that their theoretical error estimates match what happens in real computer simulations.

Technical Summary: Quadrature Error Estimates on Non-Matching Grids in a Fictitious Domain Framework for Fluid–Structure Interaction Problems

Problem Statement
The paper addresses the numerical discretization of fluid-structure interaction (FSI) problems using a fictitious domain formulation based on distributed Lagrange multipliers. In this framework, the fluid and solid are discretized using two completely independent, non-matching meshes. The solid is described in a Lagrangian framework on a reference domain $B$ , mapped to its current position within a fixed Eulerian fluid domain $\Omega$ . The kinematic coupling between the fluid velocity and the solid deformation is enforced via a Lagrange multiplier $\lambda$ defined on the solid reference domain.

A critical computational challenge arises in assembling the coupling (interface) matrix. This matrix requires integrating the product of fluid basis functions (composed with the solid mapping) and solid basis functions over the solid elements. Since these functions are defined on non-matching grids, exact integration necessitates computing the geometric intersection between the fluid and mapped solid meshes to construct a composite quadrature rule. This process is computationally expensive, particularly as the coupling term must be assembled at every time step. Alternatively, an approximate approach integrates directly over the solid elements using standard quadrature rules without mesh intersection, introducing a quadrature error. The paper investigates the theoretical validity and error estimates associated with this approximate integration strategy.

Methodology
The authors analyze a steady-state Stokes-based FSI model, though they note their results are representative of more general nonlinear cases. The variational formulation involves finding the fluid velocity $u$ , pressure $p$ , solid deformation map $X$ , and Lagrange multiplier $\lambda$ satisfying a saddle-point system.

The study focuses on two specific choices for the bilinear form $c(\cdot, \cdot)$ used to define the coupling:

$c_0$ (Dual pairing): Based on the $L^2(B)$ inner product, where the multiplier space is the dual of $H^1(B)$ .
$c_1$ (Scalar product): Based on the $H^1(B)$ inner product, where the multiplier space is $H^1(B)$ itself.

The methodology proceeds through the following steps:

Discretization: Finite element spaces are introduced for all variables. The fluid uses the Bercovier–Pironneau element ( $P1$ -iso- $P2/P1$ ), while the solid mapping and multiplier use continuous piecewise linear elements.
Approximate Integration: The authors define discrete bilinear forms where the coupling terms are computed using standard quadrature rules on the solid mesh, ignoring the fact that the composed fluid basis functions are piecewise polynomials with discontinuities relative to the solid mesh.
Abstract Error Analysis: Following the framework of inf-sup conditions, the authors derive a general error estimate (Theorem 2) that bounds the difference between the exact discrete solution and the solution obtained with inexact integration. This bound depends on the quadrature error functionals.
Quadrature Error Estimation: The core technical contribution involves proving specific estimates for the quadrature error of the coupling terms. The authors utilize the fact that the composed function $v_h(X)$ is continuous but piecewise linear on the solid elements, leading to specific regularity properties in fractional Sobolev spaces ( $H^{1+s}$ ). They employ interpolation theory and inverse inequalities to bound the error.
Stability Proof: The authors prove that the discrete problem remains well-posed (satisfies the discrete inf-sup conditions) even when the coupling term is constructed approximately, provided specific conditions on the mesh sizes are met.

Key Contributions and Results

Well-Posedness with Approximate Coupling: The paper proves that the discrete problem is stable and well-posed even when the coupling matrix is assembled using approximate quadrature, without requiring mesh intersection.
Quadrature Error Estimates:
- For the $L^2$ -based coupling ( $c_0$ ), the quadrature error is bounded by $O(h_B^{3/2} |\log h_B|)$ . The authors show that the approximate method yields optimal convergence rates provided the solid mesh size $h_B \to 0$ .
- For the $H^1$ -based coupling ( $c_1$ ), the error analysis is more complex due to the gradient term. The error bound includes terms dependent on the ratio of mesh sizes. Specifically, optimal convergence requires not only $h_B \to 0$ but also that the ratio $h_B/h_\Omega \to 0$ (where $h_\Omega$ is the fluid mesh size). If $h_B/h_\Omega$ is kept constant, the method may lose optimality.
Inverse Inequalities: The authors derive and utilize specific inverse inequalities to handle the norms of the Lagrange multiplier in the error estimates, particularly for the dual norm case.
Numerical Validation: Four numerical tests are presented to validate the theoretical estimates:
- Test 1 & 2: Compare convergence rates for $c_0$ and $c_1$ under different mesh refinement strategies (constant ratio vs. decreasing ratio). Results confirm that $c_0$ is robust with approximate integration, while $c_1$ requires $h_B/h_\Omega \to 0$ to maintain accuracy.
- Test 3: Uses a solution with a jump in the velocity gradient across the interface (typical of FSI), confirming the expected suboptimal convergence rate of $O(h^{1/2})$ due to solution singularities.
- Test 4: Investigates higher-order elements ( $P2/P1$ ), showing that while $c_0$ remains robust, $c_1$ still requires exact integration or specific mesh ratios to achieve optimal rates.

Significance and Claims
The paper claims to provide a rigorous theoretical justification for using approximate integration in fictitious domain FSI methods, a practice often adopted for computational efficiency. The authors demonstrate that:

The "inexact" approach does not compromise the stability of the numerical scheme.
The $L^2$ -based coupling ( $c_0$ ) is particularly well-suited for approximate integration, as it does not require the fluid and solid meshes to be refined in a specific ratio to maintain accuracy.
The $H^1$ -based coupling ( $c_1$ ), while potentially offering different stability properties, imposes stricter mesh requirements ( $h_B/h_\Omega \to 0$ ) to avoid quadrature errors dominating the solution error.

The authors conclude that their theoretical results confirm the experimental observations from their previous work [12, 13], providing a solid mathematical foundation for choosing between exact (mesh-intersecting) and approximate (non-intersecting) assembly techniques based on the specific coupling form and mesh configuration.