Exponential stability of the linearized viscous Saint-Venant equations using a quadratic Lyapunov function

This paper establishes the exponential stability of the linearized viscous Saint-Venant equations in the $L^2$ norm by constructing an explicit diagonal quadratic Lyapunov function and deriving sufficient conditions on boundary parameters for small viscosities.

The Gist

Imagine you are managing a long, straight canal filled with water. Your goal is to keep the water level and flow speed steady, even when wind, rain, or a sudden gate opening causes ripples and waves. This is the real-world problem the Saint-Venant equations try to solve. They are like a rulebook for how water moves in rivers and canals.

However, the standard rulebook assumes water is "perfect"—it flows without any internal friction, like a ghost moving through air. In reality, water is sticky; it has viscosity (internal friction), like honey or thick syrup. This stickiness changes how waves behave, especially when they hit the walls of the canal.

This paper asks a simple but difficult question: If we add this "stickiness" (viscosity) to our math model, can we still guarantee that the water will calm down quickly and return to a steady state?

Here is the breakdown of their discovery, using everyday analogies:

1. The Problem: The "Perfect" vs. The "Real"

In the past, mathematicians had a great tool to prove the water would calm down: a Lyapunov function. Think of this function as a "Stability Scorecard."

• If the Scorecard goes down over time, the system is stable (the water is calming down).
• If it goes up, the system is unstable (the waves are getting wilder).

For "perfect" (non-viscous) water, scientists had a specific Scorecard that worked perfectly. It was like a generic thermostat that kept a room at a perfect temperature.

2. The Twist: Viscosity Breaks the Old Tool

The authors tried to use that same old Scorecard for "sticky" (viscous) water. It failed.

• The Analogy: Imagine you have a key that opens a door perfectly. But then, someone adds a layer of thick grease to the lock. The old key no longer fits; it jams.
• The Discovery: The authors proved that when viscosity is present, the old Scorecard (which mixed the water depth and speed together in a complex way) is mathematically impossible to use. They showed that for sticky water, the Scorecard must be "diagonal."
• What does "diagonal" mean? Imagine your Scorecard has two separate columns: one for "Water Depth" and one for "Water Speed." The old tool mixed them up (Depth + Speed). The new tool says: "We must measure Depth and Speed separately to get an accurate reading." This is a major mathematical shift.

3. The Solution: Building a New Scorecard

Once they realized the old tool was broken, they built a new, custom Scorecard specifically for sticky water.

• They designed this new tool to be "diagonal" (keeping depth and speed separate).
• They then checked if this new tool could prove the water would calm down.
• The Catch: It only works if the "stickiness" (viscosity) is small. If the water is too thick (like molasses), the math gets too messy. But for real-world rivers (which are only slightly sticky), it works perfectly.

4. The Boundary Conditions: The "Doors" of the Canal

To prove the water calms down, you also need to know what happens at the ends of the canal (the boundaries).

• The authors figured out exactly how to control the "doors" at the start and end of the canal.
• They found specific rules for how to adjust the water gates based on the water level and speed.
• The Result: If you follow these specific rules, the "Stability Scorecard" will always go down. This means any disturbance (a wave, a splash) will shrink exponentially fast. The system doesn't just stop wobbling; it snaps back to calmness very quickly.

5. Why This Matters

• Realism: Most engineering models ignore viscosity because it's hard to calculate. This paper shows that ignoring it can lead to using the wrong "Stability Scorecard," which might give a false sense of security.
• Safety: If you are building a dam or managing a flood control system, you need to know the water will return to normal quickly after a storm. This paper gives engineers a new, more accurate way to prove that safety.
• Robustness: The authors note that if this math works for the "linearized" (simplified) version, it's very likely to work for the messy, real-world non-linear version too.

Summary

Think of this paper as a mechanic realizing that the old engine tuning guide doesn't work for cars with a new type of fuel (viscosity). They proved the old guide is broken, designed a new, specialized guide that treats the engine parts separately, and showed that as long as the fuel isn't too thick, the car will still run smoothly and return to a steady speed after hitting a bump.

In short: They found the right mathematical "thermostat" for sticky water, ensuring that rivers and canals can be controlled safely and predictably.

Technical Summary

Here is a detailed technical summary of the paper "Exponential stability of the linearized viscous Saint-Venant equations using a quadratic Lyapunov function" by Amaury Hayat and Nathan Lichtlé.

1. Problem Statement

The paper addresses the exponential stability of the viscous Saint-Venant equations, which model one-dimensional open-channel flow (e.g., rivers, canals) with viscosity.

• Context: The standard Saint-Venant equations are a system of hyperbolic partial differential equations (PDEs). While their stability has been extensively studied (often using Lyapunov functions), real-world fluids possess viscosity.
• The Challenge: Adding a viscosity term (derived from the Navier-Stokes equations) transforms the system into a second-order parabolic-hyperbolic system. This introduces higher-order derivatives and singular behavior as the viscosity coefficient $\mu \to 0$.
• Objective: The authors aim to prove that the linearized viscous Saint-Venant system around a steady state is exponentially stable in the $L^2$ norm, provided the viscosity is sufficiently small and specific boundary conditions are met. They specifically investigate whether existing quadratic Lyapunov functions (effective for inviscid cases) remain valid or require modification.

2. Mathematical Model

The authors consider the viscous Saint-Venant system derived in [17]:

1. Mass Conservation: $H_t + (HV)_x = 0$
2. Momentum Conservation: $V_t + V V_x + gH_x + \frac{f(H,V)}{H} - \frac{4\mu(HV_x)_x}{H} = 0$

Where:

• $H(t,x)$ is water depth, $V(t,x)$ is horizontal velocity.
• $\mu > 0$ is the viscosity coefficient.
• $f(H,V)$ is a modified friction term.
• The system is linearized around a steady state $(H^*, V^*)$ in a subcritical (fluvial) regime ($gH^* > (V^*)^2$).
• The linearized system is a second-order PDE system in matrix form:
  $$A \mathbf{y}_{xx} + \mathbf{y}_t + B(x)\mathbf{y}_x + C(x)\mathbf{y} = 0$$
  where $\mathbf{y} = (h, v)^T$ represents perturbations in depth and velocity.

3. Methodology

The core methodology relies on the construction of an explicit quadratic Lyapunov function to demonstrate exponential decay.

A. Analysis of Lyapunov Function Structure

A critical theoretical finding is that for the viscous system, a valid quadratic Lyapunov function $W = \int \mathbf{y}^T Q(x) \mathbf{y} dx$ must be diagonal in physical coordinates ($Q$ must be a diagonal matrix).

• Contrast with Inviscid Cases: Previous works (e.g., [26], [27]) used Lyapunov functions with cross-terms (off-diagonal elements in $Q$) that worked for inviscid systems. The authors prove (Proposition 3.1) that these cross-terms cause the time derivative of the Lyapunov function to become indefinite when viscosity is introduced.
• Implication: The authors must construct a new Lyapunov function that is diagonal, specifically of the form $Q = \text{diag}(q_1, q_2)$.

B. Construction of the Lyapunov Function

The authors propose a specific diagonal matrix $Q(x)$ dependent on the steady state and viscosity:
$$Q = \text{diag}\left( g + \mu \tilde{q}_1, \quad H^* + \mu \tilde{q}_2 \right)$$
The terms $\tilde{q}_1$ and $\tilde{q}_2$ are carefully chosen to ensure:

1. Symmetry: The product $QB$ is symmetric (a requirement for the energy method).
2. Positivity: $Q$ remains positive definite for small $\mu$.
3. Convergence: As $\mu \to 0$, the function recovers the known stable Lyapunov function for the inviscid case (from [5]).

C. Stability Analysis

The proof involves analyzing the time derivative $\dot{W}$ along the system trajectories:
$$\dot{W} = \int_0^L \mathbf{y}^T \left( \gamma Q - (QC + (QC)^T) - Q_{xx}A + (QB)_x \right) \mathbf{y} \, dx + \text{Boundary Terms}$$
The authors decompose this into:

1. Integral Term ($I$): They prove that for sufficiently small $\mu$, the matrix inside the integral is negative definite, ensuring the integral term contributes to decay. This relies on asymptotic estimates of the steady state derivatives ($V^*_x, V^*_{xx}$) as $\mu \to 0$.
2. Boundary Term ($B$): They derive explicit constraints on the boundary feedback coefficients ($b_0, b_1, c_1$) such that the boundary terms are non-positive.

4. Key Contributions and Results

1. Diagonal Constraint on Lyapunov Functions

The paper establishes a fundamental limitation: Standard Lyapunov functions with cross-terms fail for viscous Saint-Venant equations. The authors prove that for the linearized viscous system, the Lyapunov matrix $Q$ must be diagonal in physical coordinates to ensure stability. This explains why previous results for inviscid systems cannot be directly extended.

2. Explicit Sufficient Conditions for Stability

The authors derive explicit conditions on the boundary control parameters ($b_0, b_1, c_1$) to guarantee exponential stability for small $\mu$.

• Subcritical Condition: The steady state must satisfy $gH^*(0) < (2+\sqrt{2})(V^*(0))^2$. This is a stricter condition than the standard subcritical flow ($gH > V^2$) required for the inviscid case, highlighting the sensitivity introduced by viscosity.
• Boundary Coefficients:
  • $b_0$ and $b_1$ must lie within specific intervals defined by the steady state values at the boundaries.
  • $c_1$ (the coefficient for the velocity gradient feedback at the outlet) must be chosen within a specific range $(c_1^-, c_1^+)$.

3. Exponential Stability in $L^2$

Under the derived conditions, the authors prove that the linearized system is exponentially stable in the $L^2$ norm. Specifically, there exist constants $C, \gamma > 0$ such that:
$$\|(h(t), v(t))\|_{L^2} \leq C e^{-\gamma t} \|(h_0, v_0)\|_{L^2}$$
This holds for any viscosity $\mu$ sufficiently small ($\mu \in (0, \mu^\circ)$).

4. Numerical Validation

Section 5 presents numerical simulations using an Implicit-Explicit (IMEX) scheme. The results confirm the theoretical predictions, showing exponential decay of the $L^2$ norm for various small viscosity values ($\mu = 0.001, 0.0001$, etc.).

5. Significance and Future Work

• Robustness to Viscosity: The work bridges the gap between idealized hyperbolic models and realistic viscous fluid dynamics, proving that stability can be maintained with appropriate boundary control even when viscosity is present.
• Practical Control Design: The explicit formulas for boundary coefficients ($b_0, b_1, c_1$) provide engineers with concrete guidelines for designing feedback controllers for open-channel flows (e.g., irrigation canals, flood control) where viscosity cannot be ignored.
• Limitations: The current result is for the linearized system. While quadratic Lyapunov functions are often robust to nonlinearities, proving local exponential stability for the full nonlinear viscous system remains an open question.
• Future Directions: The authors suggest extending these results to sloped channels (where gravity terms dominate) and investigating the nonlinear stability of the full system.

In summary, this paper provides a rigorous mathematical framework for stabilizing viscous open-channel flows, correcting previous assumptions about Lyapunov function structures and offering explicit, implementable control laws.