🔢 mathematics

On the application of the SCD semismooth* Newton method to solving Stokes problem with stick-slip boundary conditions

This paper proposes and validates an efficient SCD semismooth* Newton method with a globalization technique for numerically solving the 3D Stokes problem with Navier-Tresca stick-slip boundary conditions, which is formulated as a variational inequality of the second kind and approximated via the mixed finite element method.

Original authors: V. Arzt, P. Beremlijski, H. Gfrerer, J. V. Outrata

Published 2026-04-01

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: V. Arzt, P. Beremlijski, H. Gfrerer, J. V. Outrata

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 water flows through a complex pipe system, like the blood vessels in a human brain or a simple cube of water. In physics, this is usually described by the Stokes equations. Think of these equations as the "rules of the road" for slow-moving, thick fluids (like honey or blood) that don't swirl wildly.

Usually, we assume that if water hits a wall, it stops dead in its tracks (the "no-slip" rule). But in the real world, sometimes water slides a little bit along the wall, or it sticks and then slips. This is called stick-slip behavior. It's like a car tire: sometimes it grips the road perfectly (stick), and sometimes it skids (slip). The paper focuses on a specific, tricky version of this rule called Navier-Tresca, where the fluid only slips if the force pushing it is strong enough to overcome a certain "friction threshold."

The Problem: A Mathematical Puzzle

The authors are trying to solve a massive mathematical puzzle to figure out exactly how this fluid moves.

The Challenge: The "stick-slip" rule is nonsmooth. In math terms, it's like a graph with a sharp corner or a cliff. If you try to use standard tools (like a smooth ramp) to climb this cliff, they fail or get stuck.
The Old Way: Previous methods tried to "smooth out" the cliff or turn the problem into a different type of puzzle (a dual problem) to make it easier. This is like trying to flatten a mountain to build a road; it works, but you lose some of the mountain's original shape and it takes a lot of extra work.

The Solution: The "SCD Semismooth* Newton" Method

The authors introduce a new, super-powered tool called the SCD semismooth Newton method* (let's call it the SSSN for short).

Here is an analogy to understand how it works:

The Old Newton Method: Imagine you are trying to find the bottom of a valley in the dark. The classic Newton method is like taking a step, checking the slope, and taking a giant leap in the direction that goes down. It works great if you are already close to the bottom, but if you are far away or the ground is bumpy, you might overshoot and fall off a cliff.
The "Semismooth" Twist: The SSSN method is like having a smart flashlight that can see the "sharp corners" of the terrain. Instead of trying to smooth the mountain, it understands that the ground is jagged. It knows exactly how to step on those jagged rocks without slipping.
The "SCD" Secret Sauce: The "SCD" (Subspace Containing Derivative) part is like having a GPS that knows the local geometry. When the method encounters a "stick" (where the fluid is stuck) or a "slip" (where it's moving), it instantly switches its internal map to the correct version for that specific spot. It doesn't need to guess; it calculates the perfect direction to jump next, even on the sharpest corners.

What Did They Do?

The team took this high-tech mathematical tool and applied it to a 3D simulation of fluid flow with these tricky stick-slip walls.

The Test: They ran simulations on two shapes: a simple cube and a complex model of a Cerebral Aneurysm (a bulge in a brain artery).
The Result: The SSSN method was incredibly fast and efficient.
- It didn't matter how they started the simulation (even with a bad guess); the method found its way to the correct answer.
- It solved the problem with fewer steps than older methods.
- It handled the "sharp corners" of the stick-slip rules perfectly without needing to smooth them out first.

Why Does This Matter?

Think of this as upgrading from a bicycle to a high-performance off-road vehicle.

Old Methods (Bicycle): Good for smooth roads (simple physics), but struggle on rough, rocky terrain (complex stick-slip boundaries). They might get stuck or take a very long, winding path.
New Method (Off-road Vehicle): Built specifically to handle the rocks and cliffs. It gets to the destination faster, uses less fuel (computer time), and can handle the toughest terrain without breaking a sweat.

The Bottom Line

This paper presents a new, smarter way to simulate how fluids move when they interact with walls that are sometimes sticky and sometimes slippery. By using a specialized mathematical "GPS" (the SSSN method), they can solve these complex 3D problems much faster and more accurately than before. This could help engineers design better medical devices, improve oil extraction, or understand blood flow in the human body with greater precision.

1. Problem Statement

The paper addresses the numerical solution of the 3D Stokes problem for incompressible, viscous Newtonian fluids under Navier-Tresca stick-slip boundary conditions.

Physical Context: Unlike standard no-slip conditions (where velocity is zero at the wall), real-world scenarios often involve slip. The Navier-Tresca condition models a threshold behavior:
- Stick state: If the tangential stress is below a threshold $g$ , the fluid adheres to the boundary (zero tangential velocity).
- Slip state: If the tangential stress exceeds $g$ , the fluid slips, with the slip velocity proportional to the stress.
Mathematical Formulation:
- The problem is formulated as a variational inequality of the second kind coupled with an equality constraint (incompressibility).
- The weak formulation involves finding velocity $u$ and pressure $p$ such that:
  $a_\kappa(u, v-u) + b(p, v-u) + j(v) - j(u) \geq l(v-u)$
  $b(q, u) = 0$
  where $j(v)$ represents the non-smooth friction term involving the $L^1$ -norm of the tangential velocity.
Discretization: The problem is discretized using the Mixed Finite Element Method (FEM) with P1-bubble/P1 (MINI) elements. This choice ensures stability (satisfying the inf-sup condition) while maintaining a lower degree of freedom compared to P2-P1 elements.
Resulting System: The discretization leads to a Generalized Equation (GE) of the form:
$0 \in F(x) = H(x) + \partial q(x)$
where $H$ is an affine linear operator derived from the Stokes equations, and $\partial q$ is the subdifferential of a non-smooth convex function representing the stick-slip constraints.

2. Methodology: SCD Semismooth* Newton (SSSN)

The authors propose a novel numerical approach using the SCD (Subspace Containing Derivative) semismooth Newton method*, rather than traditional smoothing techniques or dual formulations (like active set methods).

Core Concept: The method treats the non-smooth problem directly as a Generalized Equation without reformulating it into a system of smooth equations. It relies on the SCD property of set-valued mappings.
SCD Mappings: The method utilizes the fact that the subdifferential of a convex, lower semicontinuous function is an SCD mapping. This allows for the construction of a linearization (approximation) of the generalized equation using subspaces containing the derivative, rather than a single Jacobian matrix.
Algorithm Steps:
1. Approximation Step: Compute a proximal point $z^{(k)}$ using the proximal mapping of the non-smooth part $q$ . This effectively identifies the active set (stick vs. slip) locally.
2. Newton Step: Construct a linear system based on the SCD derivative (specifically, selecting matrices $P$ $P$ and $W$ $W$ that represent the subspace containing the derivative).
  - For the stick-slip problem, the matrices $P$ and $W$ are block-diagonal.
  - Slip state: $P$ and $W$ are derived from the smooth part of the subdifferential.
  - Stick state: $P=0$ and $W=I$ (identity).
3. Globalization: To ensure convergence from arbitrary starting points, the method employs a hybrid strategy. It attempts a Newton step with a line search. If the Newton step fails to reduce the residual sufficiently, it switches to a Douglas-Rachford step (a globally convergent method for monotone operators).
Implementation Details:
- The linear systems are solved using GMRES.
- To improve efficiency, the authors eliminate the internal velocity degrees of freedom (bubble nodes) to form a reduced system involving only boundary nodes and pressure, solved via a Schur complement approach.

3. Key Contributions

Direct Application to Stick-Slip: The paper successfully applies the SCD semismooth* Newton method to the Stokes problem with Navier-Tresca conditions, avoiding the need for dual variable reformulation or smoothing of the non-smooth terms.
Theoretical Guarantees: The authors prove that the subdifferential of the stick-slip functional satisfies the SCD property and is semismooth*. They demonstrate that the resulting linearized system is non-singular, ensuring local superlinear convergence of the algorithm.
Efficient Implementation: The derivation of specific block-diagonal structures for the SCD matrices ( $P$ and $W$ ) allows for efficient assembly and solution of the Newton system.
Globalization Strategy: The integration of the Douglas-Rachford method ensures robustness, allowing the solver to converge even with poor initial guesses.

4. Numerical Results

The authors tested the SSSN method on two 3D examples:

Example 1 (Unit Cube):
- Setup: A unit cube with varying slip limits ( $g=0, 5, 10$ ) and adhesive coefficients ( $\kappa=5$ ).
- Findings: The number of SSSN iterations was independent of the mesh size (ranging from ~700 to ~19,000 nodes).
- Performance: The method converged in 3–8 iterations regardless of the mesh refinement. CPU time scaled linearly with the problem size, demonstrating the efficiency of the linear solver and the mesh-independence of the Newton convergence rate.
- Scenarios: Successfully handled pure slip ( $g=0$ ), mixed stick-slip ( $g=5$ ), and pure adhesion ( $g=10$ ).
Example 2 (Cerebral Aneurysm):
- Setup: A complex 3D geometry representing a cerebral aneurysm with $\approx 13,800$ nodes and $\approx 73,000$ tetrahedra.
- Findings: The method converged for various slip bounds ( $g=0, 5, 10$ ).
- Comparison: The authors compared SSSN against a standard semismooth Newton method (from reference [9]). The SSSN method showed slightly better convergence rates and, crucially, did not require casting the problem into an equation form (avoiding the complexity of dual formulations).
- Behavior: The algorithm correctly identified the transition between stick and slip nodes during iterations, with the number of active slip nodes stabilizing as convergence approached.

5. Significance and Conclusion

Robustness: The SCD semismooth* Newton method provides a robust, mesh-independent solver for complex fluid-structure interaction problems involving non-smooth boundary conditions.
Versatility: The approach is not limited to Stokes flow; the authors note it can be adapted for Navier-Stokes and unsteady problems with minor implementation changes.
Advantage over Alternatives: By avoiding the reformulation of the problem into a dual system or smoothing the non-smooth terms, the method preserves the original problem structure, potentially reducing computational overhead and improving accuracy in capturing the exact stick-slip interface.
Future Work: The success of this method suggests it is a viable candidate for large-scale simulations in biomechanics (e.g., blood flow in vessels) and industrial fluid dynamics where boundary slip is significant.

In summary, the paper presents a mathematically rigorous and computationally efficient framework for solving 3D Stokes problems with stick-slip boundaries, leveraging advanced variational analysis (SCD theory) to achieve superlinear convergence without the drawbacks of traditional smoothing or dualization techniques.