🔢 mathematics

Formalizing the Classical Isoperimetric Inequality in the Two-Dimensional Case

This paper presents a formal verification of the classical isoperimetric inequality in the plane using the Lean 4 proof assistant and Mathlib, following Adolf Hurwitz's analytic approach to prove that among all simple closed curves of a given perimeter, the circle uniquely maximizes the enclosed area.

Original authors: Miraj Samarakkody

Published 2026-03-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Miraj Samarakkody

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

The Big Picture: The "Fence and Field" Problem

Imagine you have a fixed length of rope (let's say 100 meters). You want to lay this rope on the ground to enclose a patch of grass. You can make the rope into a square, a triangle, a wiggly snake-shape, or a perfect circle.

The Question: Which shape gives you the biggest patch of grass inside?

The Answer: A circle. No matter how you twist your rope, you can never get more grass than you would with a perfect circle. This is the Isoperimetric Inequality. It's a rule of the universe: for a given perimeter, the circle is the most efficient shape.

The Mission: Teaching a Robot to Prove It

The author of this paper, Miraj Samarakkody, didn't just want to say the circle is the best; they wanted to prove it to a computer so perfectly that the computer could never find a single logical error.

They used a tool called Lean 4, which is like a super-strict robot lawyer. In normal math, a human proof might skip a tiny step because "it's obvious." But Lean 4 is like a lawyer who says, "I don't care if it's obvious. Show me the exact rule that allows you to skip that step, or I reject your proof."

The goal was to take a famous proof by a mathematician named Adolf Hurwitz (from 1902) and translate it into code that the robot lawyer would accept.

The Strategy: The "Musical" Approach

Hurwitz's proof is clever because it doesn't use geometry (like drawing shapes). Instead, it uses music.

Imagine the curve of your rope as a song.

The Song: Any wiggly line can be broken down into a mix of simple musical notes (sine and cosine waves). This is called a Fourier Series. A complex shape is just a song made of many different notes played together.
The Volume (Parseval's Theorem): There's a rule in music theory that says the total "loudness" (energy) of the song is equal to the sum of the loudness of each individual note.
The Comparison (Wirtinger's Inequality): Hurwitz found a way to compare the "loudness" of the shape's position to the "loudness" of how fast the shape is changing (its speed). He proved that the speed is always "louder" than the position, unless the shape is a perfect circle.

The Two-Phase Journey

The paper describes the work in two main phases:

Phase 1: Building the Musical Library

Before they could prove the fence problem, they had to teach the robot the rules of music theory.

Orthogonality: They proved that different musical notes (like a high C and a low G) don't interfere with each other when you add them up.
Uniform Convergence: They proved that if you add enough notes, the song stops wiggling and settles into a smooth tune.
Differentiation: They taught the robot how to take the "derivative" (the speed) of a song made of infinite notes.

Analogy: This is like building a dictionary and a grammar book for the robot before asking it to write a poem.

Phase 2: The Proof Itself

Once the robot knew the music rules, they applied them to the rope:

Re-wrapping the Rope: They took the rope (length $L$ ) and stretched it out to fit exactly into a 24-hour clock cycle (from 0 to $2\pi$ ).
The Area Formula: They used a "shoelace formula" (a way to calculate the area of a polygon by tracing its edges) to turn the shape into a math equation.
The Trade-Off: They used a basic math rule (AM-GM) to say, "The area is limited by the sum of the square of the position and the square of the speed."
The Final Blow: They used the "Music Rules" (Wirtinger's Inequality) to show that the "speed part" of the equation is always bigger than the "position part," unless the shape is a circle.
The Result: They calculated that the maximum area you can get is $L^2 / 4\pi$ . If you try to make a square or a triangle, you fall short of this number.

The Hiccups: Why Was This Hard?

You might think, "It's just math, why is it hard for a computer?" The paper explains that computers are terrible at "common sense."

The "Infinite" Problem: In human math, we often swap a sum (adding numbers) and an integral (finding area) without thinking twice. The robot says, "No! You must prove it's safe to swap them, or the universe might break." The author had to write specific code to prove it was safe.
The "Index" Problem: Mathematicians sometimes count starting at 0, sometimes at 1. The robot gets confused if you switch. The author had to write code to translate between "Counting like a human" and "Counting like a robot."
The "Smoothness" Problem: The rope has to be perfectly smooth (no sharp corners) for the math to work. The author had to explicitly tell the robot, "Assume the rope is smooth," because the robot wouldn't guess that.

The Takeaway

This paper is a victory for certainty.

For Mathematicians: It proves that a 100-year-old proof is rock-solid. Even if we missed a tiny detail in 1902, the computer checked every single step and found no errors.
For the Future: It shows that we can use computers to verify complex, graduate-level math. It's like having a "spell-checker" for the most important laws of the universe.

In short: The author taught a robot to listen to the "music" of a shape, proved that a circle is the only shape that sings the perfect note, and did it in a language the robot couldn't argue with.

Based on the provided paper, here is a detailed technical summary of the formalization of the classical isoperimetric inequality in the two-dimensional case.

1. Problem Statement

The paper addresses the Isoperimetric Inequality in the Euclidean plane. The theorem states that for any simple closed curve $C$ with perimeter $L$ and enclosed area $A$ , the inequality $L^2 \geq 4\pi A$ holds. Equality is achieved if and only if the curve is a circle.

While the result is a classical theorem with known proofs, the paper focuses on the formal verification of a specific analytic proof by Adolf Hurwitz (1902) using the Lean 4 proof assistant and its mathematical library, Mathlib. The goal is to eliminate the "gaps" often found in textbook proofs (such as unjustified interchanges of limits and integrals) by providing a machine-checked, step-by-step logical derivation.

2. Methodology

The formalization follows Hurwitz's analytic approach, which reduces the geometric problem to an inequality involving Fourier series. The work is divided into two distinct phases:

Phase I: Fourier-Analytic Foundations

Before addressing the curve, the authors formalized the necessary tools from harmonic analysis within Mathlib:

Orthogonality Relations: Formalized the orthogonality of trigonometric functions ( $\sin(nx), \cos(nx)$ ) over $[-\pi, \pi]$ .
Convergence Theory: Established uniform convergence of Fourier partial sums using the Weierstrass M-test. This was crucial for ensuring that properties like continuity and integrability are preserved in the limit.
Term-by-Term Differentiation: Proved conditions under which a Fourier series can be differentiated term-by-term, requiring the summability of $n(|a_n| + |b_n|)$ .
Parseval's Theorem: Formalized the identity relating the $L^2$ norm of a function to the $\ell^2$ norm of its Fourier coefficients.
Wirtinger's Inequality: Derived this inequality (relating the $L^2$ norm of a zero-mean periodic function to its derivative) directly from Parseval's theorem.

Phase II: The Geometric Proof (Hurwitz's Argument)

The authors modeled a simple closed $C^1$ curve $\gamma$ in $\mathbb{R}^2$ and executed the proof in five logical steps:

Reparametrization: The curve was reparametrized from arc-length $s \in [0, L]$ to an angular parameter $\theta \in [0, 2\pi]$ to facilitate Fourier analysis.
Area Formula: The enclosed area $A$ was expressed using the Shoelace formula (Green's Theorem): $A = \frac{1}{2} \int_0^L (x y' - y x') dt$ . After reparametrization and integration by parts, this simplified to $A = \int_0^{2\pi} f(\theta)g'(\theta) d\theta$ , where $f$ and $g$ are the coordinate functions.
AM-GM Inequality: Applied the inequality $2fg' \leq f^2 + (g')^2$ to bound the area: $A \leq \frac{1}{2} \int_0^{2\pi} (f^2 + (g')^2) d\theta$ .
Wirtinger's Application: Assuming the centroid is at the origin (zero mean for $f$ ), Wirtinger's inequality ( $\int f^2 \leq \int (f')^2$ ) was applied. This upgraded the bound to $A \leq \frac{1}{2} \int_0^{2\pi} ((f')^2 + (g')^2) d\theta$ .
Arc-Length Constraint: Using the fact that the curve is parametrized by arc length, the authors showed that $(f')^2 + (g')^2 = (L/2\pi)^2$ . Substituting this constant into the integral yielded the final bound: $A \leq \frac{L^2}{4\pi}$ .

3. Key Contributions

First Complete Formalization of Hurwitz's Proof: While partial formalizations exist in other systems (Coq, Isabelle), this is a complete, machine-verified proof of Hurwitz's specific analytic approach in Lean 4.
Expansion of Mathlib: The project extended Mathlib's capabilities in classical real analysis, specifically:
- Implementing classical Fourier series with explicit coefficient sequences (rather than just abstract $L^2$ theory).
- Formalizing the Weierstrass M-test and term-by-term differentiation theorems for trigonometric series.
- Bridging the gap between abstract Euclidean space definitions (EuclideanSpace) and coordinate-wise calculus.
Handling Technical Nuances: The authors successfully navigated complex formalization challenges, including:
- Interchanging Sums and Integrals: Explicitly invoking the Dominated Convergence Theorem via MeasureTheory.integral_tsum_of_summable_integral_norm.
- Indexing Conventions: Resolving mismatches between natural number indexing ( $\mathbb{N}$ ) and positive natural number indexing ( $\mathbb{N}^+$ ) used in Fourier coefficients.
- Differentiation of Projections: Correctly composing Fréchet derivatives of projections with the curve's derivative to handle coordinate functions in $\mathbb{R}^2$ .

4. Results

The formalization successfully verified the main theorem:
$\text{For any simple closed } C^1 \text{ curve } \gamma, \quad \text{Area}(\gamma) \leq \frac{\text{Length}(\gamma)^2}{4\pi}$
The proof is available in the public repository IsoperimetricInequality on GitHub. The codebase is structured into two main files:

IsoperimetricInequality.Basic: Contains the Fourier analysis foundations (Orthogonality, Parseval, Wirtinger).
IsoperimetricInequality.Adolf Hurwitz proof: Contains the geometric derivation and the final inequality.

5. Significance

Certainty in Classical Analysis: The work demonstrates that even well-established classical theorems contain subtle technical details (regularity conditions, convergence justifications) that are often glossed over in human-readable proofs but are mandatory for formal verification.
Benchmark for Lean 4: This project serves as a significant case study for the maturity of Lean 4 in handling graduate-level classical analysis, including integration, differentiation, and infinite series.
Foundation for Future Work: By formalizing the Fourier-analytic machinery, the authors have created reusable components for future formalizations of spectral theory, variational calculus, and other geometric inequalities.
Open Problems Identified: The paper highlights that while the inequality is proven, the equality case (characterizing the circle as the unique maximizer) and the derivation of Parseval's identity directly from the curve's coordinates (rather than assuming it as a hypothesis) remain open challenges for future work.

In summary, this paper represents a rigorous, machine-checked reconstruction of a classic geometric proof, showcasing the power of modern proof assistants to validate the logical underpinnings of mathematical analysis.