🔢 mathematics

Non-Conforming Structure Preserving Finite Element Method for Doubly Diffusive Flows on Bounded Lipschitz Domains

This paper establishes a well-posedness and regularity analysis for stationary doubly diffusive flows with temperature-dependent viscosity on bounded Lipschitz domains and introduces a fully non-conforming finite element method using Crouzeix-Raviart elements that guarantees locally divergence-free velocity fields, ensures unique discrete solutions, and provides proven a priori error estimates validated through numerical tests.

Original authors: Jai Tushar, Arbaz Khan, Manil T. Mohan

Published 2026-03-27

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Jai Tushar, Arbaz Khan, Manil T. Mohan

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 complex fluid moves inside a container. But this isn't just water; it's a special fluid where two things are driving the movement: heat (like hot air rising) and salt concentration (like sugar dissolving in tea). This is called "doubly diffusive flow."

Now, imagine the container has a weird shape (like an L-shape or a room with a jagged corner), and the fluid gets thicker or thinner depending on how hot it is. This makes the math incredibly difficult.

This paper is about building a new, smarter computer simulation tool to solve these tricky problems. Here is the breakdown using simple analogies:

1. The Problem: The "Jigsaw Puzzle" that Doesn't Fit

In the past, scientists used a method called "Finite Elements" to simulate these flows. Think of this like trying to cover a floor with square tiles.

The Old Way: They used tiles that had to fit perfectly together at the edges. If the floor had a weird shape (a "Lipschitz domain," which is math-speak for a room with sharp corners or jagged walls), the tiles either didn't fit well, or the math got so complicated that the computer couldn't guarantee a single, unique answer. It was like trying to solve a puzzle where the pieces might fit in two different ways, leaving you confused about the final picture.
The New Way: The authors propose a "Non-Conforming" method. Imagine using tiles that are allowed to overlap slightly or have gaps, as long as the average height at the edge matches up. It's more flexible.

2. The Secret Weapon: The "Perfectly Balanced" Flow

The biggest breakthrough in this paper is that their new method ensures the fluid is locally divergence-free.

The Analogy: Imagine a crowd of people moving through a hallway. "Divergence-free" means that for every person who enters a small room, exactly one person must leave. No one gets created or destroyed inside the room.
Why it matters: In the old methods, the computer sometimes "created" or "lost" tiny amounts of fluid in small pockets due to rounding errors. This new method uses Crouzeix-Raviart elements (a specific type of mathematical tile) that act like a strict bouncer at a club: "If you come in, you must go out." This keeps the simulation physically realistic and stable.

3. The "Uniqueness" Guarantee

In math, sometimes a problem can have multiple answers (like a ball sitting at the bottom of a valley that has two dips).

The Old Problem: Previous methods couldn't prove that the computer would find the one correct answer. It might get stuck in a "fake" solution.
The New Solution: Because the new method is so strict about keeping the fluid balanced (divergence-free), the authors proved mathematically that there is only one correct answer. It's like proving that if you follow these specific rules, there is only one way to finish the puzzle.

4. Handling the "Weird Shapes"

Most simulations work great on perfect squares or circles. But real-world things (like blood vessels, oil reservoirs, or ocean currents) have jagged edges.

The Achievement: The authors proved their method works even on "Lipschitz domains" (rooms with sharp corners). They showed that even if the solution gets a little "rough" near the corners (like a fabric fraying), their method can still handle it and give a good answer. They didn't need to assume the room was perfectly smooth.

5. The "Recipe" for Accuracy

The paper also provides a "recipe" for how accurate the answer will be.

The Analogy: If you double the number of tiles you use (make the tiles smaller), the picture becomes sharper. The authors calculated exactly how much sharper. They found that even on rough shapes, the error drops at a predictable rate.
The Test: They ran the simulation on a "porous cavity" (like water flowing through a sponge). They compared their results to famous benchmarks (the "gold standard" of previous research). Their method matched the experts' data very closely, proving it works in the real world.

Summary

Think of this paper as inventing a new, more flexible type of Lego brick for building fluid simulations.

It's flexible: It works on weird, jagged shapes where old bricks failed.
It's strict: It ensures no fluid is magically created or lost (conservation of mass).
It's reliable: It guarantees that the computer finds the one true answer, not a fake one.
It's proven: It works on both smooth rooms and jagged corners, and it matches real-world experiments.

This is a big step forward for engineers and scientists who need to model complex flows in oil extraction, climate science, or food processing, especially when the shapes involved aren't perfect.

1. Problem Statement

The paper investigates the mathematical and numerical analysis of a stationary doubly diffusive flow model in two and three dimensions on bounded Lipschitz domains ( $\Omega \subset \mathbb{R}^d, d=2,3$ ). The model couples the incompressible Navier-Stokes equations with temperature and concentration-dependent advection-diffusion equations.

The governing system (Equation 1.1) is:
$\begin{cases} K^{-1}u + (u \cdot \nabla)u - \text{div}(\nu(T)\nabla u) + \nabla p = F(y) & \text{in } \Omega, \\ \text{div } u = 0 & \text{in } \Omega, \\ -\text{div}(D\nabla y) + (u \cdot \nabla)y = 0 & \text{in } \Omega, \\ y = y_D, \quad u = 0 & \text{on } \Gamma, \end{cases}$
where:

$u$ and $p$ are fluid velocity and pressure.
$y = (T, S)^\top$ represents temperature ( $T$ ) and concentration ( $S$ ).
$\nu(T)$ is a temperature-dependent viscosity (Lipschitz continuous and bounded).
$K$ is a permeability matrix (modeling Darcy/Brinkman effects).
$D$ is a diffusion matrix (allowing for cross-diffusion/Soret and Dufour effects).
$F(y)$ is a buoyancy force term.

Key Challenge: The analysis is conducted under minimal regularity assumptions on the domain geometry (Lipschitz boundaries, potentially non-convex) and data, rather than assuming smooth boundaries or high regularity of the solution.

2. Methodology

A. Mathematical Analysis (Continuous Level)

Well-posedness: The authors establish the existence of weak solutions using the Faedo-Galerkin approximation technique.
Regularity Analysis: A significant portion of the work is dedicated to proving regularity results for weak solutions on Lipschitz domains.
- They utilize Stokes regularity results (Lemma 3.7) and elliptic regularity (Lemma 3.8) on specific classes of domains ( $O_2(\mathbb{R}^2)$ and $O_3(\mathbb{R}^3)$ ).
- They prove that the weak solution $(u, p, y)$ belongs to $H^{3/2+\delta}(\Omega)$ for $\delta \in (0, 1/2)$ , even with variable viscosity and nonlinear coupling.
- For convex domains, they show the solution achieves $H^2(\Omega)$ regularity.
Uniqueness: Uniqueness of the weak solution is proven under a small data assumption (or sufficiently large viscosity/diffusion coefficients). This is achieved by deriving a "weak-strong uniqueness" result using the minimal regularity established, avoiding the need for $W^{1,\infty}$ assumptions on velocity used in previous literature.

B. Numerical Discretization

The paper proposes a fully non-conforming finite element method based on the lowest-order Crouzeix-Raviart (CR) elements for velocity and concentration, coupled with piecewise constant elements for pressure.

Velocity/Concentration Space ( $V_h^{cr}$ ): Continuous at the midpoints of interior edges/faces but discontinuous across them.
Pressure Space ( $Q_h$ ): Piecewise constants ( $P_0$ ).
Structure Preservation: The chosen pair $(V_h^{cr}, Q_h)$ ensures that the discrete velocity field is locally exactly divergence-free ( $\text{div}_h u_h = 0$ element-wise). This is crucial for preserving conservation laws and energy stability.
Convective Terms: Handled via an upwind discretization to ensure stability and coercivity of the convective operator.
Discrete Kernel: The discrete kernel of the divergence operator is exactly the space of discrete divergence-free functions, which simplifies the analysis.

3. Key Contributions

Minimal Regularity Framework: Unlike previous works that required smooth domains or high regularity data to prove uniqueness, this paper establishes existence and uniqueness for Lipschitz domains with minimal data regularity.
Uniqueness of Discrete Solutions: A major theoretical hurdle in non-conforming methods (specifically for Navier-Stokes) is proving the uniqueness of the discrete solution. The authors successfully prove uniqueness for the discrete system under the same minimal regularity assumptions used for the continuous problem. This is achieved by leveraging the locally divergence-free property of the CR elements, which eliminates certain penalty-dependent norms that usually obstruct uniqueness proofs in other schemes.
A Priori Error Estimates: The paper derives optimal a priori error estimates for velocity, concentration, and pressure.
- For general Lipschitz domains: Convergence rate of $O(h^{1/2+\delta})$ .
- For convex domains: Optimal convergence rate of $O(h)$ .
Structure-Preserving Scheme: The method guarantees local mass conservation (divergence-free velocity) without requiring $H(\text{div})$ -conforming elements (like Raviart-Thomas), which are often more complex to implement for coupled systems.

4. Numerical Results

The authors validate their theoretical findings through numerical experiments using the FEniCS library:

Accuracy Tests (Convex Domain):
- Verified convergence rates in Flow, Stokes, and Darcy regimes.
- Results confirmed the theoretical $O(h)$ convergence for convex domains (Remark 6.2).
- The discrete velocity was confirmed to be divergence-free up to machine precision ( $\approx 10^{-15}$ ).
Accuracy Tests (Non-Convex Domain):
- Tested on an L-shaped domain to simulate re-entrant corners.
- Observed convergence rates of $O(h^{1/2+\delta})$ (specifically $O(h^{2/3})$ for the manufactured solution), matching the theoretical predictions for reduced regularity.
Benchmark Test (Porous Cavity):
- Simulated the Soret and Dufour effects in a porous cavity (Darcy regime).
- Compared Nusselt ($Nu$) and Sherwood ($Sh$) numbers against existing literature ([13], [46], [47]).
- The proposed method showed excellent agreement with benchmark data for $Nu$ and reasonable agreement for $Sh$, with discrepancies attributed to the difficulty of resolving thin boundary layers with lowest-order elements on uniform grids.

5. Significance and Impact

Theoretical Advancement: The paper bridges a gap in the mathematical theory of doubly diffusive flows by extending well-posedness and uniqueness results to non-smooth (Lipschitz) domains, which are common in real-world applications (e.g., porous media, geological formations).
Computational Efficiency: By using Crouzeix-Raviart elements, the method avoids the high degrees of freedom associated with $H(\text{div})$ -conforming elements while still maintaining the critical property of local mass conservation.
Robustness: The scheme is shown to be robust across different physical regimes (from pure fluid flow to Darcy flow) and geometric complexities (convex and non-convex domains).
Uniqueness Guarantee: The proof of discrete uniqueness under minimal assumptions is a significant contribution, as many existing non-conforming schemes for Navier-Stokes-type problems struggle to guarantee unique discrete solutions without strong regularity assumptions.

In summary, this work provides a rigorous mathematical foundation and a practical, structure-preserving numerical algorithm for simulating complex, temperature-dependent doubly diffusive flows in realistic, non-smooth geometries.