Linear Kelvin Wave Predictions in the $z\to 0$ Limit

This paper resolves the unbounded wave energy issue in the $z \to 0$ limit of linear Kelvin wave theory by introducing a modified kernel with elliptic spanwise integration and a fast contour-deformation evaluator, enabling physically consistent and computationally efficient predictions for flat-ship applications.

The Gist

Imagine you are trying to predict the wake left behind by a boat moving through water. For engineers, this is crucial for designing faster, more fuel-efficient ships.

For decades, scientists have used a "mathematical shortcut" called Linear Wave Theory to do this. It's like using a simple sketch instead of a photorealistic 3D rendering: it's incredibly fast and captures the main shape of the waves, making it perfect for testing thousands of ship designs quickly.

However, there's a major glitch in this shortcut.

The Problem: The "Infinite Wave" Glitch

When a ship's hull is very flat and sits right on the surface of the water (a common design for high-speed boats), the math breaks down.

Think of the mathematical tool used to predict waves as a flashlight.

• When the light source is deep underwater, the beam is focused and manageable.
• But as the light source moves up to the very surface of the water, the beam doesn't just get brighter; it explodes. The math predicts that the waves would become infinitely tall and contain infinite energy.

In the real world, waves don't do that. But in the computer model, this "infinite energy" causes the simulation to crash or produce nonsense. To fix this, engineers usually have to add "band-aids"—arbitrary rules or manual tweaks—to stop the numbers from blowing up. This makes the model less reliable and harder to trust.

The Solution: The "Fuzzy Flashlight"

The author of this paper, Gabriel Weymouth, found a way to fix the math without needing any band-aids.

Instead of treating the ship as a single, sharp point (like a laser pointer), he treated the ship's hull as a fuzzy, spread-out line (like a soft glow stick).

Here is the analogy:

• The Old Way (Point Source): Imagine trying to balance a heavy weight on the tip of a needle. It's unstable and falls over (the math explodes).
• The New Way (Line Integration): Now, imagine spreading that same weight out over a wide, flat board. The pressure is distributed evenly, and the system is stable.

By mathematically "smearing" the ship's influence across its width (using a specific shape called an elliptic distribution, which looks like a flattened oval), the author discovered that the "infinite" waves naturally smooth out. The energy becomes finite and realistic, even when the ship is sitting exactly on the water's surface.

The Super-Fast Calculator

Even with the math fixed, calculating these waves is usually slow. It's like trying to count every single grain of sand on a beach to measure the tide.

The author also built a super-fast calculator (a new algorithm).

• Old Method: Trying to count every grain of sand one by one. (Takes hours).
• New Method: Using a satellite to estimate the sand density and a clever trick to skip the boring parts. (Takes milliseconds).

This new calculator is 10,000 to 100,000 times faster than previous methods. It's so fast that it could potentially be used for real-time ship control or to train AI models to design ships instantly.

What Did They Find?

Using this new, stable, and fast method, the author simulated a flat boat and found some interesting things:

1. No More Magic: The waves behaved exactly as physics should, with no infinite spikes.
2. The "Rooster Tail": The simulation correctly showed a specific splash pattern (a "rooster tail") forming right behind the boat, which matches what you see in real life.
3. Wider is Calmer: They found that as the boat gets wider, the waves actually get smaller relative to the boat's size. This is because the waves generated by the left and right sides of the boat cancel each other out (like noise-canceling headphones), a detail older, broken models missed.

Why Does This Matter?

This paper gives engineers a reliable, fast, and free tool to design ships.

• No more guessing: You don't need to manually tweak the math to make it work.
• Real-time design: Because it's so fast, you could theoretically design a ship while watching the waves form on your screen in real-time.
• Better AI: Since the math is now clean and consistent, it's perfect for training Artificial Intelligence to become a master ship designer.

In short, the author took a broken, explosive mathematical tool, fixed the core flaw by changing how the ship is "seen" by the math, and built a rocket engine to make the calculations lightning fast.

Technical Summary

1. Problem Statement

The paper addresses a fundamental numerical and physical inconsistency in linear potential flow theory regarding the Kelvin Green's function for a translating point source.

• The Singularity: As the source depth ($z$) approaches the free surface ($z \to 0^-$), the wave elevation spectrum $S_\zeta(k)$ diverges. The spectrum peaks at a wavenumber $k^* \sim 1/|z|$ with an amplitude $\sim 1/|z|$.
• Consequences: This divergence implies unbounded wave energy, making the downstream wake unresolvable. It causes severe numerical difficulties for standard quadrature methods and creates physical inconsistencies for surface-piercing bodies (flat ships) where the source effectively lies on $z=0$.
• Current Limitations: Existing methods, such as the Neumann-Michell theory, avoid this singularity by using implicit iterative schemes that sidestep direct evaluation on $z=0$, sacrificing the efficiency and directness of explicit kernel representations.

2. Methodology

The author proposes a two-pronged approach combining analytical regularization and numerical acceleration:

A. Analytical Regularization (The Modified Kernel)

Instead of using a point-source kernel, the paper adopts a flat-ship model where the source strength is distributed along the spanwise direction ($y$) on the free surface ($z=0$).

• Elliptic Distribution: By imposing a constant-downwash boundary condition (representing a shallow-draft ship with a small pitch angle $\alpha$), the source strength distribution $q(y)$ is derived to be elliptic:
  $$q(y_p) = q_0 \sqrt{1 - (y_p/b)^2}$$
  where $b$ is the half-beam.
• Regularization Mechanism: The paper analytically integrates the point-source Green's function against this elliptic distribution. This integration transforms the amplitude function of the wavelike term.
  • The point-source amplitude is constant ($A \equiv 1$).
  • The line-integrated amplitude involves a Bessel function: $A_b(t) = \pi J_1(b k_y) / (b k_y)$.
• Resulting Spectrum: This modification filters out high-wavenumber content. The wave elevation spectrum for the line-integrated kernel decays as $S_\zeta(k) \sim k^{-3/2}$, which is finite and integrable even at $z=0$, completely eliminating the unbounded energy of the point source.

B. Numerical Acceleration (Partitioned Contour Deformation)

Direct numerical integration of the wavelike kernel is inefficient or non-convergent near $z=0$ due to rapid oscillations and non-analytic phase functions.

• Partitioning Strategy: The integral is split into two regions:
  1. Non-oscillatory intervals: Around stationary points (where the phase derivative is zero), standard adaptive Gauss-Kronrod quadrature is used on the real axis.
  2. Semi-infinite tails: Complex contour deformation (steepest descent) is applied. The integration path is shifted into the complex plane to induce exponential decay.
• Handling Non-Analyticity: Unlike previous works dealing with analytic phases, the Kelvin phase $g(x,y,t) = (x+yt)\sqrt{1+t^2}$ has branch points at $t = \pm i$. The method carefully selects the branch of the square root to ensure phase continuity along the deformed contours.
• Hankel Decomposition: For the line-integrated kernel, the Bessel function in the amplitude is decomposed into Hankel functions to isolate oscillatory terms, allowing the contour deformation to be applied effectively to the shifted phases.

3. Key Contributions

1. Resolution of the $z \to 0$ Singularity: The paper demonstrates that the singularity is not intrinsic to linear potential flow but is an artifact of the point-source assumption. By using an elliptic spanwise distribution derived from flat-ship theory, the wave energy becomes finite and physically consistent at the free surface.
2. Analytical Derivation of the Regularized Kernel: The derivation of the elliptic distribution and its resulting $k^{-3/2}$ spectral decay provides a rigorous theoretical basis for flat-ship predictions without empirical tuning.
3. High-Speed Evaluator: A novel partitioned contour-deformation algorithm is developed that achieves a $10^4$ to $10^5$ speedup over direct quadrature while maintaining relative errors below $10^{-6}$. This makes real-time evaluation feasible.
4. Open-Source Implementation: A Julia implementation is provided to ensure reproducibility and facilitate adoption in design and machine learning applications.

4. Results

• Wave Patterns: Simulations of a rectangular planform on $z=0$ show physically consistent wave fields. The model captures the "rooster-tail" wake behind the stern and correctly predicts the destructive interference of diverging waves as the beam increases.
• Wave Resistance Trends:
  • Length ($L$): The resistance coefficient exhibits classical constructive and destructive interference patterns based on the bow-stern distance.
  • Beam ($b$): Contrary to classical thin-ship theory (which predicts resistance $\propto b^2$), the new model predicts a monotonic decrease in wave resistance as the beam increases ($C_W \sim b^{-3}$ for large $b$). This is attributed to increased spanwise destructive interference of diverging waves, encoded in the Bessel function amplitude.
• Numerical Performance: The method handles the merging of transverse and diverging wave ridges and the $z \to 0$ limit gracefully, with error envelopes insensitive to the observation point.

5. Significance

• Design and Optimization: The ability to compute wave resistance and free-surface elevation directly on $z=0$ without empirical damping or iterative schemes makes this method ideal for design optimization and real-time control of high-speed vessels.
• Machine Learning: The computational efficiency ($10^5$ speedup) and robustness make this kernel suitable as a physics-informed backbone for surrogate modeling and machine learning applications, where thousands of evaluations are required.
• Theoretical Clarity: The work resolves a decades-old issue in hydrodynamics, clarifying that the "singularity" is a modeling artifact of point sources rather than a fundamental flaw in linear theory, and provides a rigorous path forward for surface-piercing body analysis.