🔢 mathematics

Cubature from rational approximation

This paper presents a numerical method for constructing high-precision cubature rules for area integrals over planar domains by reducing the problem to a contour integral via the Cauchy–Green identity and approximating boundary values with rational functions using the AAA algorithm, where the resulting poles and residues serve as nodes and weights with rigorous error control.

Original authors: Gentian Zavalani

Published 2026-07-21

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Gentian Zavalani

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a cartographer trying to measure the total "stuff" inside a mysterious, irregularly shaped island. In the world of mathematics, this "stuff" is an area integral, a way of adding up values across a surface. Usually, to do this, you have to break the island into millions of tiny, manageable tiles, measure each one, and sum them up. It's like trying to count every grain of sand on a beach by picking them up one by one. But what if the island is made of a special kind of magic material—analytic functions—that behaves perfectly smoothly? In this case, mathematicians have long known a secret shortcut: you don't need to look inside the island at all. If you know the rules of the game (specifically, a classical identity called the Cauchy–Green formula), you can figure out the total amount of "stuff" just by walking along the shoreline and listening to the waves.

This paper, titled "Cubature from Rational Approximation," takes that shoreline shortcut and supercharges it. The authors, led by Gentian Zavalani, ask a bold question: Can we turn this shoreline listening into a precise, automated machine that finds the exact answer for any smooth shape? They propose a method that treats the shoreline data like a puzzle. Instead of using a rigid grid of tiles, they use a flexible, intelligent "net" made of rational functions (mathematical expressions that look like fractions). This net is cast over the boundary of the shape, and where the net catches the most interesting "snags" (mathematical poles) inside the island, those spots become the measurement points. The paper suggests that by analyzing how well this net fits the shoreline, we can predict the total area with incredible accuracy, often reaching the limits of computer precision. It's a way of turning a complex 2D problem into a clever 1D boundary trick, revealing hidden skeletons inside shapes that mathematicians have only dreamed of.

The Shoreline Detective

Let's say you want to calculate the total volume of water in a strangely shaped swimming pool, but you can't dive in. You can only walk around the edge. In the past, if the pool was a perfect circle or a rectangle, you could use a simple formula. But for a pool shaped like a starfish or a jagged rock, the math gets messy. Usually, you'd have to chop the pool into a grid of tiny squares, measure the depth of each, and add them up. This is slow and prone to errors if the grid isn't fine enough.

The authors of this paper found a way to skip the grid entirely. They start with a clever mathematical trick called the Cauchy–Green identity. Think of this as a magic rule that says: "If you know the shape of the edge, you can calculate the whole inside." Specifically, if you have a function that describes the "weight" of the water at every point, you can turn the problem of measuring the whole pool into a problem of measuring just the edge.

But here's the catch: to use this trick, you need to know a specific value along the edge, which is like knowing the "shadow" of the water's weight. For simple shapes, this shadow is easy to write down. For complex shapes, it's a nightmare. This is where the paper's main innovation comes in. The authors use a computer algorithm called AAA (which stands for Adaptive Antoulas–Anderson, though you can just think of it as a super-smart pattern matcher) to approximate this shadow.

Imagine the AAA algorithm as a detective trying to draw a picture of a suspect based only on a blurry photo of their shadow. The detective tries to fit a "rational function"—a fancy type of mathematical curve made of fractions—to the shadow data along the edge. As the detective tries to fit the curve better and better, something magical happens: the curve develops "kinks" or "snags" inside the shape. These snags are called poles.

The Magic of the Snags

In the world of this paper, these poles are the heroes. The authors discovered that the locations where these mathematical snags appear inside the shape are exactly where you should place your measurement points. It's as if the algorithm is saying, "Hey, the math tells me the most important information is hiding right here, so let's put a sensor there."

The paper shows that if you take these pole locations and assign them specific weights (numbers that tell you how much to count each point), you get a cubature rule. This is just a fancy name for a formula that adds up a few points to get the total area. The beauty is that you don't need to know the shape of the pool in advance; you just need the edge data. The algorithm figures out where to look inside.

The authors tested this on several shapes:

The Circle: The algorithm correctly found that the only important point is the center, reproducing the classic rule that the average value on a circle is the value at the center.
The Ellipse: For an oval shape, the poles lined up perfectly along the "focal segment" (the line connecting the two foci), matching a known mathematical reduction.
The Starfish: For a wobbly, star-shaped domain, the poles didn't just scatter randomly. They formed a beautiful, five-armed skeleton inside the shape. The authors call this a "mother body," a hidden analytic skeleton that seems to hold the shape together.
The Square: Even for a square with sharp corners, the method worked. The poles clustered tightly near the corners and along the diagonals, converging to the answer with "root-exponential" speed. This means that as you add more points, the error drops incredibly fast, much faster than standard methods.

When Things Get Tricky

The paper is careful to point out that this isn't a magic wand for everything. It works best when the function being measured is "analytic," meaning it's smooth and doesn't have sudden jumps or breaks. If the function is messy or discontinuous, this method might not work. Also, the weights (the numbers you multiply the points by) don't always have to be positive. Sometimes, you might get a negative weight, which sounds weird (like subtracting water), but mathematically it canc out errors perfectly.

One interesting scenario the paper explores is when the shape has a "double pole" (a mathematical singularity that acts like two points very close together). In these cases, the algorithm might find two points that are practically on top of each other with huge, opposite weights. The authors suggest that instead of treating them as two separate points, you should treat them as a single "derivative" point, which captures the slope of the function at that spot. This allows the method to handle shapes that require more complex rules, like those found in "quadrature domains."

The Verdict

The paper doesn't claim to have solved every problem in the universe. It explicitly states that it doesn't provide a general formula for non-smooth functions, and the error bounds are only "rigorous" once you've proven the error on the boundary is small enough. However, the results are impressive. In their simulations, the method recovered known exact answers for circles and ellipses to machine precision (the limit of how accurate a computer can be). For the starfish and square, it achieved errors as low as $10^{-15}$ or $10^{-16}$ , which is essentially zero for all practical purposes.

The authors also found that the method works for calculating "volume potentials," which are like measuring the gravitational pull of the pool from a point outside or inside it. Even when the point is very close to the edge (making the math very difficult), the method held up, provided the edge data was sampled finely enough.

In short, this paper presents a new, automated way to turn a 2D area problem into a 1D boundary problem. By using a smart algorithm to fit a curve to the edge, it discovers the hidden "skeleton" of the shape and uses that to calculate the total area with astonishing precision. It's a bit like finding the secret code to a treasure map just by tracing the coastline, revealing that the treasure isn't buried randomly, but lies along a specific, elegant path waiting to be found.

Technical Summary: Cubature from Rational Approximation

Problem Statement
The paper addresses the numerical construction of cubature rules for area integrals of analytic functions over planar domains $\Omega \subset \mathbb{C}$ with rectifiable Jordan boundaries. The goal is to approximate integrals of the form:
$I = \iint_{\Omega} f(z) w(z, \bar{z}) \, dA$
where $f$ is analytic in a neighborhood of $\bar{\Omega}$ and $w$ is a weight function. The challenge lies in generating a discrete sum of the form $\sum c_k f(z_k)$ that approximates this integral with high accuracy, without requiring a priori knowledge of the domain's internal geometry or the specific integrand $f$ , relying instead on boundary data.

Methodology
The proposed method, termed the AAA cubature rule, transforms the area integral into a boundary approximation problem through the following steps:

Boundary Reduction via Cauchy–Green Identity:
The method begins by selecting a $\bar{\partial}$ -antiderivative $W$ such that $\partial W / \partial \bar{z} = w$ . Using the Cauchy–Green identity, the area integral is converted into a contour integral over the boundary $\gamma = \partial \Omega$ :
$\iint_{\Omega} f(z) w(z, \bar{z}) \, dA = \frac{1}{2i} \oint_{\gamma} f(z) W(z, \bar{z}) \, dz$
For pure area integrals ( $w \equiv 1$ ), $W = \bar{z}$ .
Rational Approximation on the Boundary:
The boundary data $\omega(t) = W(z(t), \overline{z(t)})$ is approximated by a rational function $\rho(z)$ with free poles using the AAA algorithm (Adaptive Antialiasing Algorithm). The approximation minimizes the error on a sampled set of boundary points.
$\rho(z) = \sum_{z_k \in \Omega} \frac{c_k}{z - z_k} + \sum_{z_k \notin \Omega} \frac{d_k}{z - z_k} + p(z)$
where $p(z)$ is a polynomial.
Extraction of Cubature Rule:
By applying residue calculus to the contour integral of $f(z)\rho(z)$ , the method derives a cubature rule:
$I_n = \pi \sum_{z_k \in \Omega} c_k f(z_k)$
The nodes are the poles of $\rho$ located inside $\Omega$ , and the weights are $\pi c_k$ . Poles outside $\Omega$ and the polynomial part integrate to zero against analytic functions and are discarded for the final rule.
Error Control and Validation:
The error is controlled by the quality of the rational fit on the boundary. Theorem 2.1 provides an a posteriori error bound:
$|I - I_n| \leq \frac{1}{2} |\gamma| \varepsilon \max_{z \in \gamma} |f(z)|$
where $\varepsilon = \max_{z \in \gamma} |\rho(z) - W(z, \bar{z})|$ . The continuous supremum $\varepsilon$ is estimated via independent validation grids.
Handling Confluence and Singularities:
When poles cluster (confluent pairs) to represent derivative functionals (as in quadrature domains with meromorphic Schwarz functions), the raw weights are unstable. The method evaluates such clusters using local moments $\mu_j = \sum \lambda_k (z_k - z_C)^j$ rather than individual weights.

Key Contributions and Results

Recovery of Classical Rules: The method successfully recovers known exact identities to machine precision, including the disk mean-value rule and the focal-segment reduction for ellipses (Gegenbauer quadrature).
Quadrature Domains: For domains where the Schwarz function is meromorphic (quadrature domains), the algorithm automatically discovers the exact finite quadrature identities. It handles both separated nodes and confluent pairs (representing derivatives) by identifying the underlying pole structure.
Root-Exponential Convergence: For domains with corners (e.g., a square), the method exhibits root-exponential convergence ( $O(e^{-C\sqrt{n}})$ ), characteristic of "lightning approximation," with nodes clustering along diagonals and corners.
Potential Theory Connections: The interior poles trace analytic skeletons identified tentatively as "mother bodies" of potential theory. For volume potentials (logarithmic and Cauchy kernels), the method evaluates integrals for targets both inside and outside the domain using only boundary data, revealing image-point structures without explicit imposition.
Dual Interpretation: The paper establishes a dual reading where the cubature rule corresponds to the exact integral of a rational interpolant to the integrand, analogous to the one-dimensional interpolatory viewpoint.

Significance and Claims
The paper claims to provide a direct numerical construction that generates weighted cubature rules from boundary data alone, without requiring new transforms or internal mesh generation. Its significance lies in:

Generality: It applies to any rectifiable Jordan boundary and various weight functions, provided a $\bar{\partial}$ -antiderivative exists.
Adaptivity: Unlike classical rules that rely on fixed geometric models (e.g., focal segments), the AAA rule adapts to the actual boundary data, maintaining accuracy even when the geometric model is perturbed.
Rigorous Bounds: It offers a rigorous error estimate based entirely on the boundary residual, independent of the interior behavior of the antiderivative.

The authors remain modest regarding the theoretical underpinnings of the node geometry, noting that while the numerical evidence strongly suggests a connection to mother bodies and image points, these are not yet proven to be the exact objects of potential theory. Similarly, while the method handles non-positive and complex weights effectively, it does not claim to be a general-purpose formula for non-analytic integrands. The work serves as a forward counterpart to inverse shape-from-moments problems, reconstructing approximate quadrature identities from boundary data.

The Shoreline Detective

The Magic of the Snags

When Things Get Tricky

The Verdict

Technical Summary: Cubature from Rational Approximation

More like this