How does the binormal flow model three-dimensional vortex filament dynamics?

A local approximation standing in for the Biot-Savart integral

The object being modeled is a vortex filament: the fluid's vorticity is a vector measure supported on a curve χ(s,t) in three-dimensional space, parametrized by arc length s and evolving in time t. Outside the curve, the velocity field follows from the Biot-Savart integral, but evaluating that integral on the curve itself is obstructed by its singularity. The lectures follow the Localized Induction Approximation, which keeps only local effects of the integral; after a logarithmic renormalization in time, the filament moves in the binormal direction with speed proportional to curvature, giving χt = χs ∧ χss.

The paper is explicit that this is a purely ideal model, first appearing in Da Rios's 1906 work and rederived by Arms and Hama in 1965 as an approximation of vortex filament dynamics under the Euler equations. Its purpose is to supply a proper toy model for vortex filaments and, potentially, for turbulence theory. The authors also state a clear disadvantage: the model does not exclude dynamics absent from real fluids, since closed curves preserve length under the flow while fluids exhibit filamentation.

Whether the approximation is legitimate is the vortex filament conjecture: can one prove that Euler solutions exist whose vorticity stays concentrated along a curve evolving by binormal flow for long times? The paper reports that this remains open in the general case, and lists partial results, including the Fontelos-Vega 2023 result that for Giga-Miyakawa solutions starting from regular closed curves without self-intersections, the leading order behaves as a Lamb-Ossen vortex adapted to a curve moving by an appropriately normalized binormal flow.

Three equations describe the same filament

The model's analytical value comes from a chain of equivalences. If χ solves the binormal flow, its tangent vector T = χs satisfies Tt = T ∧ Tss, the one-dimensional Schrödinger map on the sphere S2, a structure the paper links to Landau-Lifshitz ferromagnetism theory. Separately, the Hasimoto transformation converts the binormal flow into a focusing one-dimensional cubic nonlinear Schrödinger equation.

These are not three competing models but three descriptions of one object, and each description supplies tools the others lack. The Schrödinger-map view exposes the geometric constraint that the tangent vector stays on the sphere. The NLS view turns a question about curve stability into a scattering problem for a dispersive equation. The paper stresses that the NLS solutions involved live in a low-regularity regime, so the standard conservation laws are unavailable and new tools are required.

The paper's worked example of the payoff is the self-similar corner. The authors focus on singular filaments, curves smooth except at corner-like singularities, because experimental vortices appear to be built from self-similar curves: the same picture recurs at different scales, with a circle-like shape near the vertex and helices whose axis emanates from the corner. The corresponding self-similar solutions are real analytic for positive times and V-shaped at the initial time.

A Gaussian factor fixes the corner angle

For the self-similar family, the paper gives an explicit asymptotic description. The solution χa = √t Ga(s/√t) stays within a√t of a pair of straight rays A+a s for s ≥ 0 and A−a s for s < 0, so the corner is asymptotically two rays meeting at a vertex. The vectors A±a are unit vectors, and their components obey A+a1 = A−a1 = e^(−πa²/2) while the second and third components flip sign between the two sides.

That first component carries the geometry: the paper states sin(θ/2) = A±a1 = e^(−πa²/2), where θ is the angle between A+a and −A−a. The parameter a therefore controls how sharp the corner is, and the same exponential appears in the hyperbolic analogue with the sign pattern changed. The paper also records a self-intersection threshold in terms of a: there exist 0 < a0 ≤ a1 < +∞ such that χa has no self-intersections when a < a0 and finitely many when a ≥ a1.

The asymptotic expansion also shows how the nonlinearity enters. The two independent solutions of the relevant linear equation have dispersion relation φ(ξ) = ξ² − (a²/2) log|ξ|, so the paper describes the nonlinear effect as only a logarithmic correction to free evolution. This is a statement about this self-similar construction, not a claim that nonlinearity is negligible in general.

Stability of the corner becomes a scattering question

The paper's stability claim is that the corner survives perturbation: if one starts from a filament close to the self-similar solution, the corner is still formed. The route to proving this is the Hasimoto correspondence, which translates the stability issue into a scattering problem for solutions of the cubic NLS.

The qualification matters as much as the claim. Because the relevant NLS solutions sit in a low-regularity regime, the argument cannot lean on standard conservation laws, and the paper says new tools are introduced for this purpose. The stability result should therefore be read as a statement about the specific self-similar corner solutions and their perturbations within that functional setting, not as a blanket stability theorem for arbitrary filament data.

Polygonal filaments return with rotated axes

For several corners, the paper works through polygonal initial data. Jerrard and Smets supply a global existence theorem covering such data and simulate the unit square, which appears to reappear at a certain time with its axes turned by π/4. The lectures go further: for a regular polygon with M sides, another polygon appears at times tp,q = (2π/M²)(p/q) with p and q coprime, and except for q = 2 the result is a skew polygon with Mq sides when q is odd or Mq/2 sides when q is even.

The mechanism is the Talbot effect in the associated NLS. At rational times the solution can be written as a finite superposition of translates of the initial datum, and the paper derives this by evaluating a generalized quadratic Gauss sum; the resulting polygon's adjacent-side angle is constant, and the shape is fixed by the Gauss-sum phases. The paper notes that the conjecture cos(ρ) = 2cos^(2/q)(π/M) − 1 for odd q, and 2cos^(4/q)(π/M) − 1 for even q, was formally justified elsewhere.

The axis-switching case is the q = 2 instance: the Gauss sum vanishes for half the cases, so the polygon at that time has M sides rather than 2M, which is how the paper explains the π/M rotation. The authors connect this to documented observations in non-circular jets and to a domestic smoke-cannon experiment with an equilateral triangular hole showing a flip-flop with axis switch of π/6 for M = 3, while noting that viscosity smooths the corners.

A Riemann-type function tracks a polygon corner

The trajectory of a single corner is where the model meets number theory. De la Hoz and Vega observed numerically that a mild variant of Riemann's non-differentiable function, φ(t) = Σ_{n∈Z} (e^(2πin²t) − 1)/n², closely represents the corner trajectory of polygonal vortices, and Banica and Vega later proved under certain hypotheses that the rescaled trajectory MχM converges to the plane curve φx0(t) = Σ_{n≠0} e^(2πi(n²t + nx0))/n² as M → +∞.

The paper reports that this function has since been analyzed: for rational x0 it satisfies the Frisch-Parisi multifractal formalism from turbulence theory, and for irrational x0 it has a multifractal behaviour. Fractal structure already appears at the linear level of the 1d NLS with piecewise-constant data, where rational times give finite overlaps of translates and irrational times give fractal profiles. The lectures extend this line to filaments with corners expressed as sums of Dirac distributions.

Where the toy model meets real fluid behavior

The paper is careful about the gap between the ideal flow and actual fluids. Straight lines, circles, and helices exist both in experiments and as Euler solutions, but the authors note this does not exclude non-fluid dynamics in the model. The strongest bridge reported is the Fontelos-Vega result for regular closed curves without self-intersections, and the paper states it is open whether that can be extended to more singular solutions allowing corners in the initial filament.

The experimental parallels are presented as suggestive rather than conclusive. The self-similar solutions are compared to a flow traversing a triangular wing, and axis switching is linked to observations for non-circular jets and nozzles shaped as ellipses, rectangles, squares, or equilateral triangles. The smoke-cannon experiment shows the M = 3 flip-flop, but the paper says the predicted six-sided skew polygons from the Talbot effect are not conclusive and require further evidence.

The priority the paper establishes is therefore specific: the self-similar corner, with its Gaussian angle factor e^(−πa²/2) and its stability via NLS scattering, is the best-supported link between binormal flow and observed vortex structure, while the polygonal and Talbot predictions are mathematically derived and numerically supported but experimentally open. Readers weighing the model should treat the corner results and the polygon predictions as different tiers of evidence rather than a single uniform claim.

Sources used in this answer