Isoperimetric Problems

⚡ Key Takeaways

One-line statement: Among all closed plane curves of a fixed perimeter, the circle encloses the maximum area; equivalently, for a fixed area the circle has the least perimeter. Key result in words: The isoperimetric inequality states $L^2 \geq 4\pi A$ for any simple closed curve of length $L$ enclosing area $A$ , with equality if and only if the curve is a circle. Where it applies: Geometry, calculus of variations, optimization, physics (surface tension, minimal energy), and as a foundational example in variational problems with constraints. Who/when: Posed in antiquity (Dido's problem, ~800 BC); rigorous proofs from Steiner (1838), Weierstrass, Schwarz, and Hurwitz (1901).

💡 The Core Insight

Optimizing one geometric quantity subject to a constraint on another forces a balance condition: at the optimum, the boundary has constant curvature. In the plane the only closed curve of constant curvature is the circle. Algebraically, the constraint introduces a Lagrange multiplier $\lambda$ , and the Euler–Lagrange equation reduces to "curvature = constant = $1/\lambda$ ." Geometry (symmetrization) and analysis (variational calculus, Fourier series) give two completely different routes to the same conclusion.

DefinitionSimple Closed Curve, Perimeter, Enclosed Area

A simple closed curve in $\mathbb{R}^2$ is a continuous map $\gamma:[0,L]\to\mathbb{R}^2$ with $\gamma(0)=\gamma(L)$ and no other self-intersections (a non-self-crossing loop). Parametrize by arc length $s$ , so $\gamma(s)=(x(s),y(s))$ with $\dot x^2+\dot y^2=1$ .

Perimeter (length):

1

\quad L=\int_0^L ds=\int_0^L\sqrt{\dot x(s)^2+\dot y(s)^2}\,ds.

Derivation hint: arc length is the integral of the speed $\|\gamma'\|$ ; with arc-length parametrization the integrand is identically $1$ .

Enclosed area (Green's theorem):

2

\quad A=\frac12\oint_\gamma\bigl(x\,dy-y\,dx\bigr)=\frac12\int_0^L\bigl(x\dot y-y\dot x\bigr)\,ds.

Derivation hint: apply Green's theorem to $P=-y/2,\,Q=x/2$ , giving $\iint_D(Q_x-P_y)\,dA=\iint_D 1\,dA=A$ .

TheoremIsoperimetric Inequality

For every simple closed curve in the plane with perimeter $L$ enclosing area $A$ ,

3

\quad L^2\ge 4\pi A,

with equality if and only if the curve is a circle.

TheoremDual Form

Equivalently, among all regions of fixed area $A$ , the disk minimizes perimeter:

4

\quad L\ge\sqrt{4\pi A}=2\sqrt{\pi A}.

ProofHurwitz's Fourier-Series Proof of the Isoperimetric Inequality

This is the cleanest analytic proof, due to Hurwitz (1901).

Step 1 — Normalize. By scaling, assume the perimeter is $L=2\pi$ . Reparametrize by a parameter $t\in[0,2\pi]$ proportional to arc length, so that $t=2\pi s/L=s$ . Then $x(t),y(t)$ are $2\pi$ -periodic, and the unit-speed condition gives

5

\quad \dot x^2+\dot y^2=1\quad\Longrightarrow\quad \int_0^{2\pi}\!\bigl(\dot x^2+\dot y^2\bigr)\,dt=2\pi.

Derivation hint: integrate the constant integrand $1$ over $[0,2\pi]$ .

Step 2 — Fourier expand. Write

6

\quad x(t)=\frac{a_0}2+\sum_{n\ge1}(a_n\cos nt+b_n\sin nt),\qquad \dot x(t)=\sum_{n\ge1}n(-a_n\sin nt+b_n\cos nt),

and similarly for $y$ with coefficients $c_n,d_n$ . Derivation hint: differentiate term-by-term; the constant $a_0/2$ vanishes upon differentiation.

Step 3 — Parseval. By Parseval's identity,

7

\quad \int_0^{2\pi}\!\bigl(\dot x^2+\dot y^2\bigr)\,dt=\pi\sum_{n\ge1}n^2\bigl(a_n^2+b_n^2+c_n^2+d_n^2\bigr)=2\pi,

and the area from [2] becomes

8

\quad A=\frac12\int_0^{2\pi}\!(x\dot y-y\dot x)\,dt=\pi\sum_{n\ge1}n\bigl(a_nd_n-b_nc_n\bigr).

Derivation hint: Parseval converts integrals of products of trig series into sums of coefficient products; cross terms vanish by orthogonality of $\{\cos nt,\sin nt\}$ .

Step 4 — Compare. Subtract twice [8] from [7]:

9

\quad 2\pi-2A=\pi\sum_{n\ge1}\Bigl[n^2(a_n^2+b_n^2+c_n^2+d_n^2)-2n(a_nd_n-b_nc_n)\Bigr]\ge 0.

Derivation hint: each summand equals $\pi\bigl[(na_n-d_n)^2+(nb_n+c_n)^2+(n^2-n)(c_n^2+d_n^2)\bigr]\ge0$ since $n\ge1$ makes $n^2-n\ge0$ .

Step 5 — Conclude. Thus $A\le\pi$ . Since $L=2\pi$ , this reads $A\le L^2/(4\pi)$ , i.e. $L^2\ge 4\pi A$ . Equality forces every term to vanish: $n^2-n=0$ kills all $n\ge2$ terms, leaving $a_1=d_1,\,b_1=-c_1$ for $n=1$ . Then

10

\quad x=\tfrac{a_0}2+a_1\cos t+b_1\sin t,\quad y=\tfrac{c_0}2-b_1\cos t+a_1\sin t,

which traces a circle. $\blacksquare$

∎

ProofVariational Derivation: Constant Curvature

Maximize $A$ (eq. [2]) subject to fixed $L$ (eq. [1]) via a Lagrange multiplier $\lambda$ . Form

11

\quad J=A-\lambda L=\int_0^L\Bigl[\tfrac12(x\dot y-y\dot x)-\lambda\sqrt{\dot x^2+\dot y^2}\Bigr]\,ds.

Derivation hint: the constrained extremum of $A$ subject to $L=\text{const}$ is an unconstrained extremum of $A-\lambda L$ . The Euler–Lagrange equations for $x$ and $y$ , after using arc-length parametrization, yield

12

\quad \dot y=\lambda\,\ddot x,\qquad -\dot x=\lambda\,\ddot y\ \Longrightarrow\ \kappa=\frac{\dot x\ddot y-\dot y\ddot x}{(\dot x^2+\dot y^2)^{3/2}}=\frac1\lambda=\text{const}.

Derivation hint: the signed curvature $\kappa$ collapses to the constant $1/\lambda$ ; the only simple closed plane curve with constant curvature is a circle of radius $|\lambda|$ .

∎

mathematician · Jakob Steiner | 1796–1863

A self-taught Swiss geometer who became professor at Berlin, Steiner championed synthetic (coordinate-free) geometry and disdained analysis. His 1838 attack on the isoperimetric problem introduced symmetrization: given any non-circular region, one constructs a region of the same perimeter and larger area by reflecting across a chord, proving the circle is optimal — provided an optimum exists. Weierstrass famously pointed out this existence gap, which Steiner's argument could not fill. Steiner's symmetrization nonetheless became one of the most powerful tools in geometric analysis, later refined to prove the inequality rigorously and generalized to higher dimensions.

🧠 Intuition

Think of a fixed-length elastic loop laid on a table. To enclose the most area you push it outward uniformly until it is taut everywhere equally — no dents, no bulges. A dent (concave spot) wastes perimeter while losing area, and a sharp corner concentrates "turning" inefficiently. The fully efficient shape distributes its turning uniformly: equal curvature at every point. That is exactly a circle. Soap films and bubbles "discover" this automatically because surface tension minimizes boundary energy at fixed enclosed volume — the physical analog of the dual form [4].

Figure

A non-circular loop (low area for its length) versus the optimal circle of equal perimeter

ExampleComparing a Square and a Circle of Equal Perimeter

Given: A wire of length $L=40\ \text{cm}$ .

Square: side $a=L/4=10\ \text{cm}$ , area $A_\square=a^2=100\ \text{cm}^2$ .

Circle: circumference $L=2\pi r\Rightarrow r=\dfrac{40}{2\pi}=\dfrac{20}{\pi}\approx 6.366\ \text{cm}$ .

13

\quad A_\circ=\pi r^2=\pi\left(\frac{20}{\pi}\right)^2=\frac{400}{\pi}\approx 127.3\ \text{cm}^2.

Derivation hint: substitute $r=L/(2\pi)$ into $A=\pi r^2$ to get $A=L^2/(4\pi)$ .

Check the inequality for the square: $L^2=1600$ , $4\pi A_\square=4\pi(100)\approx1256.6$ . Indeed $1600\ge1256.6$ . For the circle: $4\pi A_\circ=4\pi(400/\pi)=1600=L^2$ (equality).

Boxed answer: The circle encloses $\boxed{\approx 127.3\ \text{cm}^2}$ versus the square's $100\ \text{cm}^2$ — about 27% more area for the same wire.

ExampleThe Isoperimetric Quotient

Define the dimensionless isoperimetric quotient $Q=\dfrac{4\pi A}{L^2}\in(0,1]$ , with $Q=1$ only for the circle.

14

\quad Q=\frac{4\pi A}{L^2}.

Derivation hint: dividing inequality [3] by $L^2$ gives $4\pi A/L^2\le 1$ ; the ratio is scale-invariant.

Equilateral triangle, side $a$ : $L=3a$ , $A=\frac{\sqrt3}{4}a^2$ .

15

\quad Q_{\triangle}=\frac{4\pi\cdot\frac{\sqrt3}{4}a^2}{9a^2}=\frac{\pi\sqrt3}{9}\approx 0.6046.

Derivation hint: cancel $a^2$ ; the area of an equilateral triangle is $(\sqrt3/4)a^2$ .

Square: $Q_\square=\dfrac{4\pi a^2}{(4a)^2}=\dfrac{4\pi}{16}=\dfrac{\pi}{4}\approx 0.7854.$

Regular $n$ -gon: $Q_n=\dfrac{\pi}{n}\cot\!\frac{\pi}{n}\to 1$ as $n\to\infty$ .

Boxed answer: Triangle $\boxed{Q\approx0.605}$ , square $0.785$ , hexagon $\approx0.907$ , circle $1$ — quotient rises monotonically toward the circle.

⚠️ Common mistake

The inequality requires a simple closed curve. If self-intersections are allowed, "area" must be interpreted with signed multiplicity (winding number), and $L^2\ge4\pi A$ can fail or become vacuous. A figure-eight, for example, has cancelling signed areas. Always assume a non-self-intersecting rectifiable Jordan curve.

⚠️ Common mistake

Steiner's symmetrization proves only conditional optimality. It shows: if a maximizer exists, it must be the circle. It does not prove existence. The existence of an optimal curve must be established separately (via compactness in a suitable function space, or by the constructive Fourier proof above). Conflating "the circle beats every competitor we can build" with "the circle is the unique global max" is the classic logical gap Weierstrass exposed.

CounterexampleWhy "Constant Curvature" Alone Is Not Enough in Higher Dimensions

In the plane, constant curvature forces a circle. One might guess "constant mean curvature (CMC) closed surface ⇒ sphere" in $\mathbb{R}^3$ . Alexandrov's theorem confirms this for embedded surfaces — but if embeddedness is dropped, Wente tori (1986) are immersed CMC tori that are not spheres. Thus the naïve "constant curvature ⇒ optimum" intuition needs the topological/embeddedness hypothesis to hold up beyond two dimensions.

connection

The isoperimetric problem is the prototypical constrained optimization in the Calculus of Variations: maximize a functional ( $A$ ) subject to an integral constraint ( $L$ ), solved by Lagrange Multipliers generalized to function space (the multiplier $\lambda$ in [11]). The equality condition "constant curvature" links it to Differential Geometry of Curves via the curvature $\kappa$ . The dual minimization form connects to Minimal Surfaces and soap-film problems, while Hurwitz's proof showcases Fourier Series and Parseval's identity as a tool in geometry. The higher-dimensional generalization belongs to Geometric Measure Theory.

🔬 Where this appears

⏳ History

- ~800 BC (legend): Dido's problem — Queen Dido of Carthage is granted as much land as an oxhide can encircle; cutting it into a thin strip, she encloses a semicircle (a circle against the coastline), the folkloric origin of the problem.

- ~150 BC: Zenodorus writes On Isometric Figures, proving (rigorously by Greek standards) that the circle exceeds any polygon of equal perimeter in area, and the sphere any solid of equal surface.

- 1700s: The Bernoullis and Euler formalize the calculus of variations; isoperimetric constraints give the field its very name ("isoperimetric problems").

- 1744: Euler systematizes constrained variational problems, the technical ancestor of the Lagrange-multiplier treatment.

- 1838: Jakob Steiner publishes the symmetrization proof — elegant but assuming existence of a maximizer.

- ~1870s: Karl Weierstrass identifies the existence gap and develops rigorous foundations for the calculus of variations.

- 1884: Hermann Schwarz proves the 3D isoperimetric inequality (sphere minimizes surface for fixed volume).

- 1901: Adolf Hurwitz gives the definitive Fourier-series proof (eq. [5]–[10]), clean and rigorous.

- 20th c.: De Giorgi, Federer, Fleming (1950s–60s) establish the inequality in

\mathbb{R}^n

within geometric measure theory; Alexandrov (1958) proves embedded CMC surfaces are spheres.

open-problem

While the planar and Euclidean $n$ -dimensional cases are fully resolved, many relatives remain hard. The isoperimetric inequality on general Riemannian manifolds is known only in special cases; sharp constants depend on curvature bounds. The Cartan–Hadamard conjecture — that the Euclidean isoperimetric inequality holds in every simply connected manifold of nonpositive curvature — is proven only in dimensions $\le 4$ (and open in general). The double-bubble problem (least-area surface enclosing and separating two given volumes) was proven for equal volumes only in 2002 (Hutchings–Morgan–Ritoré–Ros); the analogous triple-bubble and higher cases are largely open. Discrete and graph isoperimetric constants (Cheeger constants) are NP-hard to compute exactly, fueling ongoing work in spectral graph theory.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also