When Waves Meet Flows Their Own Size, Only One Property Matters

A single surface wave doesn't live alone. It rides on top of whatever the ocean, or the air, or a lab tank happens to be doing underneath it. When the length of that wave is comparable to the swirling eddies it passes over — not much shorter, as textbooks assume — the two stop behaving like a background and a foreground. They become one problem. And that problem, until now, has been a mess to solve.

This paper from Université Paris-Saclay offers a way out. Basile Gallet and Alexandre Tlili have derived a pair of clean, almost elegant equations that capture how deep-water waves scatter off a background flow whose size matches the wavelength — the regime that oceanographers, atmospheric scientists, and lab experimentalists actually keep running into, and the one that the standard "scale separation" toolkit quietly fails at. The headline isn't that waves interact with flows; everyone already knew that. It's that you can now describe that interaction with a two-dimensional Schrödinger equation, integrated on the slow timescale of the flow itself, instead of grinding through a fully three-dimensional simulation resolving every individual wave crest. That's the difference between forecasting a storm with a weather app and simulating every raindrop.

The reward is a physical claim worth pausing on. When narrow-band waves pass over a full three-dimensional background flow, the only thing that matters about the flow is its near-surface vertical vorticity — the tendency of the water to spin about the vertical axis right beneath the surface. Everything else about the flow, including whether its surface is converging or diverging, is subdominant. It slides away into terms too small to care about.

The Science

The starting point is the full Euler equations for an incompressible fluid with a free surface — the same naughty system that has occupied fluid dynamicists since Bernoulli. The fluid sits in a half-space below the surface, with gravity pulling down and a uniform-pressure atmosphere above. On top of this full system ride deep-water surface gravity waves (think swell, whose wavelength is short compared to the water depth beneath it). The paper adds a background flow of whatever shape, as long as it's slow.

"Slow" is the whole game. The Froude number — the ratio of the background flow's speed to the waves' phase speed — is small, written $ϵ=U_0/\sqrt{gL}\ll 1$. That's the regime where the background flow's turnover time is much longer than the wave period, which is exactly the situation oceanographers call realistic: mesoscale and submesoscale eddies evolve over days while swell oscillates over seconds. Crucially, unlike the classic WKB ray-tracing approach, the wavelength $L$ here is comparable to the horizontal scale of the background flow. No separation of scales. That single relaxation is what makes the problem hard, and it's why this regime has been so under-studied since the isolated efforts of Phillips (1959) and Fabrikant & Raevsky (1994).

The theoretical engine is something the authors call the equivalent solvability condition method. The idea is subtle but powerful. When you do a multiple-timescale expansion of the full 3D wave problem, the equations at each order don't automatically have solutions — they're only consistent if certain "solvability conditions" hold, the wave-mean-flow analog of a resonance condition. The authors' move is to demand that their reduced, 2D equations produce the same leading-order solution and the same first solvability condition as the full 3D system. If two equations imply the same constraints, they're equivalent for the purposes that matter. This is how they turn a fully 3D free-surface problem into a 2D equation for a single complex field, and it's a deliberate generalization of the celebrated Young & Ben Jelloul (1997) reduction that has spawned decades of near-inertial wave research.

They work two cases. First, broad-band waves above a depth-invariant background flow. Here the compact variable is $\chi$, a combination of the interfacial displacement and the surface vertical velocity:

$$\chi (x,y,t)=\frac{1}{\sqrt{2}}{D}^{-\frac{1}{4}}\eta +\frac{i}{\sqrt{2}}{D}^{-\frac{3}{4}}{w}_s,$$

where $D$ is the operator that multiplies by wavenumber in spectral space. The reduced equation,

$${\partial }_t\chi +i{D}^{\frac{1}{2}}\chi +J(\Psi ,\chi )+\frac{1}{4}{D}^{-1}\{J(\Delta \Psi ,D^{-1}\chi )\}=0,$$

captures the dispersion of the waves (the $i{D}^{1/2}\chi$ term), the advection by the flow (the Jacobian $J(\Psi ,\chi )$ term), and a genuinely new nonlocal correction involving the flow's vorticity $\Delta \Psi$. The second case is narrow-band waves above a fully 3D background flow, where the field is demodulated into $M$, a slowly varying amplitude that absorbs the fast oscillation at the central frequency, $\chi =M{e}^{-it}$. This is the equation that reduces to the elegant Schrödinger form.

What They Found

The narrow-band result deserves to be stated precisely, because its simplicity is the payoff. Define the effective 2D streamfunction $\psi$ from the background flow's vertical vorticity $\zeta =(\nabla \times \mathbf{U})\cdot {\mathbf{e}}_z$ alone:

$$\psi =\left({\Delta }^{-1}+\frac{1}{4}\right)\left\{{\int }_{-\infty }^{0}2e^{2z}\zeta (x,y,z,t)dz\right\}.$$

Notice what appears in that integral: only the vertical vorticity, weighted by a decaying exponential that fades out within about a wavelength of the surface. Nothing about horizontal divergence. The paper proves, in its most technical sections, that the near-surface horizontal divergence of the flow has a subdominant effect on the waves compared to the near-surface vorticity (Gallet & Tlili, 2026). This sharpens and generalizes the earlier result of Fabrikant & Raevsky that only vertical vorticity scatters surface waves — and it gives a practical rule for the lab: want to know how a flow will scatter your waves? Measure the spin near the surface, not the stretching.

The reduced equations conserve the right things. Both exactly conserve wave action $A=⟨|\chi {|}^2⟩$, the closest thing surface waves have to a particle number, whose density behaves like the probability density of a quantum particle. And when the background flow is steady, both conserve an energy invariant, constructed from the Hermitian Hamiltonian form $i{\partial }_t\chi =\mathcal{H}\{\chi \}$. These conservation laws aren't decorative — they're what make the reduced equations usable over long integrations without slowly leaking energy.

The paper then demonstrates the equations do real work. In the weak-scattering limit they derive an analytical solution: a Gaussian wave packet hitting a patch of organized flow gets deflected by a specific, predictable angle. Scattering is resonant only when the scattered wavevector ${\mathbf{k}}_2={\mathbf{k}}_1+\mathbf{K}$ also lands on the unit circle — a geometric condition that fixes the scattering angle as $\pm {\theta }_0=\pm \pi /3$ for the isotropic test flow (

). For a disorganized patch of flow, with random vorticity over a ring of wavevectors, they integrate the reduced equation numerically. The scattered waves emerge at the predicted first scattering angles $\pm \pi /3$ and then visibly spawn secondary scattering at $\pm 2\pi /3$ — a clean, predictable cascade emerging from a chaotic scatterer (

). Even better, the incident wave packet is shown to lose a significant fraction of its initial action to the scattered field, visible in both the surface elevation and the action density

Wave packet loses significant action to scattering by disorganized flow

Label	Value
Incident action (input)	1
Action lost to scattered field	0.35

.

Why This Changes Things

Scale separation is the comfortable assumption most of wave theory leans on, and this paper is a reminder of how often reality refuses to cooperate. In the ocean, Langmuir cells — those counter-rotating subsurface rolls that line up in the wind direction and concentrate floating debris into windrows — have wavelengths comparable to their lateral extent. That is precisely the regime this paper handles, and precisely the regime ray tracing cannot. In the atmosphere, Rossby waves have wavelengths comparable to the planetary radius; there is no "much shorter wavelength" regime to fall back on. Laboratory experiments probing flow through surface-wave scattering deliberately operate where the wavelength matches the flow scale. Every one of these systems has been waiting on equations like these.

The computational payoff is not subtle. The reduced narrow-band equation evolves $M$ only on the slow advective timescale of the background flow and the small detuning of the waves, not on the fast wave period. Concretely, this means numerical timesteps much larger than the wave period. The broad-band equation, meanwhile, still contains fast wave dynamics, but that term can be integrated exactly — a standard exponential-integrator trick that again permits timesteps comparable to the flow's turnover time. Even the spatial reduction helps: a fully 3D free-surface simulation becomes a 2D field. The authors don't quote a benchmark speedup, but the combination of dropping a spatial dimension and jumping an entire timescale is the difference between an experiment that runs overnight and one that runs in seconds.

The physical insight about vorticity is the deepest takeaway. That the flow's only relevant property is its near-surface vertical vorticity is not obvious — a wave feels the surface motion, so one might guess the surface divergence matters just as much. It doesn't. The proof occupies the most technical stretch of the paper, but the consequence is beautifully simple: the scattering problem is governed by a single scalar field derived from the flow's spin, not its stretching. For an experimentalist, that's a huge simplification in what to measure and what to control.

What's Next

The work is, self-consciously, "Part I," and the title promises more. The equivalent solvability condition method is presented as a general framework, not a one-off trick — the Yamada & Thomas reconstitution method it refines has already been applied by Onuki & Fujiwara (2026) to couple surface waves and background flows, and the Paris-Saclay method offers a broader, unified route to such coupled models. The immediate open question is the two-way coupling: these equations describe how the flow scatters waves, but a coupled model must also say how the waves push back on the flow — the domain of the Craik–Leibovich equations and the emerging work of Vanneste & Young (2026). This paper arms the wave side of that coupling with a tractable equation; wiring in the back-reaction is the natural next step.

Caveats deserve honesty. The deep-water assumption means these equations don't directly apply to shallow coastal water, where the bathymetry changes the whole game. The background flow is taken to be slow — $ϵ\ll 1$ — a genuinely real oceanographic regime, but not the only one. And the narrow-band reduction, while powerful, requires waves clustered around a central frequency; the broad-band equation covers the full spectrum but only for depth-invariant flows. Each reduction trades a slice of generality for a leap in tractability, and the authors are candid about accounting for that trade.

What the method opens up, most of all, is the ability to push further into comparable-scale scattering than analysts could before. Resonant scattering of surface waves has been the quiet cousin of the celebrated acoustic and electromagnetic scattering theories — elegant equations for light and sound have existed for a century, but water waves sat out because their governing equations were too ugly. Gallet and Tlili have effectively tamed those equations into a form their physicist cousins would recognize immediately: a Schrödinger equation, with an effective potential built from the spin of the flow, and wave action playing the role of |probability|². When a field as notoriously nonlinear as water waves finally fits into a linear operator picture, the journey from "intractable" to "textbook exercise" — and from there to practical prediction — has begun.