[Paper Review] Bilinear generalized Radon transforms in the plane

Allan Greenleaf, Alex Iosevich+2 more|arXiv (Cornell University)|Apr 4, 2017

Medical Imaging Techniques and Applications5 references3 citations

TL;DR

This paper establishes sharp $L^p \times L^q \to L^r$ boundedness estimates for a class of bilinear generalized Radon transforms in the plane, defined via integration over the unit circle with rotated variables. The key contribution is identifying the optimal exponent region $Q$, a polyhedron with specific vertices, and proving restricted weak-type estimates at the critical point $(1/2,1/2,1/2)$, with applications to counting equilateral triangles in point sets.

ABSTRACT

Let $σ$ be arc-length measure on $S^1\subset \mathbb R^2$ and $Θ$ denote rotation by an angle $θ\in (0, π]$. Define a model bilinear generalized Radon transform, $$B_θ(f,g)(x)=\int_{S^1} f(x-y)g(x-Θy)\, dσ(y),$$ an analogue of the linear generalized Radon transforms of Guillemin and Sternberg \cite{GS} and Phong and Stein (e.g., \cite{PhSt91,St93}). Operators such as $B_θ$ are motivated by problems in geometric measure theory and combinatorics. For $θ

Motivation & Objective

• To establish sharp $L^p \times L^q \to L^r$ boundedness for bilinear generalized Radon transforms in $\mathbb{R}^2$.
• To characterize the optimal range of exponents $(1/p, 1/q, 1/r)$ for which the transform is bounded.
• To extend the theory to a broader class of bilinear operators defined by Dirac delta constraints and geometric phase functions.
• To apply the results to problems in geometric measure theory, particularly counting congruent configurations such as equilateral triangles in finite point sets.
• To prove that the identified exponent region $Q$ is sharp in the scale of $L^p$ spaces with $p,q,r \geq 1$.

Proposed method

• The authors define a model bilinear transform $B_\theta(f,g)(x) = \int_{S^1} f(x-y)g(x - \Theta y)\, d\sigma(y)$, where $\Theta$ is rotation by angle $\theta \in (0,\pi]$.
• They analyze the operator using techniques from Fourier restriction theory and oscillatory integral theory, particularly focusing on the rotational curvature condition of Phong and Stein.
• The proof relies on a partition of unity argument on $\mathbb{R}^{10}$, decomposing the domain to localize the analysis to regions where one of four geometric conditions (9.2)–(9.5) holds.
• Local boundedness of the kernels is established via the implicit function theorem and non-vanishing of certain Jacobian determinants (e.g., (9.8)–(9.11)), ensuring smooth pushforwards of the product kernel.
• The authors verify that the $L^2$-norm of the bilinear operator applied to characteristic functions is controlled by $|E|^{1/2}|F|^{1/2}$, leading to the sharp endpoint estimate.
• They extend the result to a general class of bilinear operators of the form $B(f,g)(x) = \iint \delta(\phi_1(x,y)-t_1)\delta(\phi_2(x,z)-t_2)\delta(\phi_3(y,z)-t_3)f(y)g(z)\psi(y,z)\,dy\,dz$ under geometric assumptions on $\phi_j$.

Experimental results

Research questions

• RQ1 What is the optimal range of exponents $(p,q,r)$ for which the bilinear generalized Radon transform $B_\theta$ is bounded from $L^p \times L^q$ to $L^r$ in $\mathbb{R}^2$?
• RQ2 How does the boundedness depend on the angle $\theta$, particularly in the degenerate case $\theta = \pi$?
• RQ3 Can the sharp exponent region $Q$ be characterized as a polyhedron with explicit vertices, and is it optimal?
• RQ4 What geometric conditions on the defining functions $\phi_j$ ensure that the general bilinear operator satisfies the same $L^p \times L^q \to L^r$ estimates?
• RQ5 Is the point $(1/2,1/2,1/2)$ an endpoint where only a restricted weak-type estimate holds, and why?

Key findings

• The operator $B_\theta$ is bounded from $L^p \times L^q$ to $L^r$ if and only if $(1/p, 1/q, 1/r)$ lies in the polyhedron $Q$ with vertices $(0,0,0)$, $(2/3,2/3,1)$, $(0,2/3,1/3)$, $(2/3,0,1/3)$, $(1,0,1)$, $(0,1,1)$, and $(1/2,1/2,1/2)$.
• At the point $(1/2,1/2,1/2)$, only a restricted weak-type estimate holds, indicating that this is a sharp endpoint not in the strong type region.
• For $\theta = \pi$, the set of admissible exponents is strictly smaller than $Q$, indicating a degeneracy in the geometric structure.
• The exponent region $Q$ is sharp in the scale of $L^p$ spaces with $p,q,r \geq 1$, meaning no larger region supports the boundedness.
• The general bilinear operator with three Dirac delta constraints satisfies the same $L^p \times L^q \to L^r$ estimates under the rotational curvature condition and the rank condition (9.7) on the Jacobian matrix.
• The boundedness is established via local analysis using the implicit function theorem and non-vanishing of specific $4 \times 4$ Jacobian determinants (e.g., (9.8)–(9.11)), ensuring smooth pushforwards of the product kernel.