Elliptic Integrals and Elliptic Functions

⚡ Key Takeaways

Elliptic integrals are integrals of the form $\int R(x,\sqrt{P(x)})\,dx$ where $P$ is a cubic or quartic polynomial and $R$ is rational; they cannot generally be expressed in elementary functions. Elliptic functions are their inverses — doubly periodic meromorphic functions on the complex plane. Key result: every elliptic integral reduces to a combination of three Legendre canonical forms, and elliptic functions (Jacobi sn/cn/dn, Weierstrass ℘) generalize trigonometric functions to two independent periods. They appear in pendulum motion, arc length of ellipses, electromagnetism, and number theory. Founded by Legendre, Abel, Jacobi (1820s–1830s).

💡 The Core Insight

Trigonometric functions arise by inverting $\int_0^x dt/\sqrt{1-t^2} = \arcsin x$ , giving the singly periodic $\sin$ . Replace the quadratic $1-t^2$ under the root by a quartic $(1-t^2)(1-k^2t^2)$ , and inverting the integral produces a function with two independent periods. The degree of the polynomial under the radical controls the number of periods: quadratic → 1 period (trig), quartic/cubic → 2 periods (elliptic). Double periodicity is the central organizing miracle of the entire theory.

DefinitionElliptic Integral

An elliptic integral is an integral of the form

1

\quad \int R\big(x,\,\sqrt{P(x)}\big)\,dx,

where $R(u,v)$ is a rational function of $u,v$ and $P(x)$ is a cubic or quartic polynomial with no repeated roots. Derivation hint: The condition deg $P\in\{3,4\}$ with distinct roots is exactly what makes $\sqrt{P}$ an algebraic function whose integral defines a genus-1 (toroidal) Riemann surface; lower degree gives genus 0 (elementary). Every such integral can be reduced (via algebraic manipulations and partial fractions) to elementary functions plus a combination of the three Legendre canonical forms. The incomplete elliptic integral of the first kind is

2

\quad F(\varphi,k)=\int_0^\varphi \frac{d\theta}{\sqrt{1-k^2\sin^2\theta}}=\int_0^{\sin\varphi}\frac{dt}{\sqrt{(1-t^2)(1-k^2t^2)}},

where $\varphi$ is the amplitude, $k\in[0,1]$ is the modulus, and $t=\sin\theta$ converts trigonometric to algebraic ("Jacobi") form. Derivation hint: Substitute $t=\sin\theta$ , $dt=\cos\theta\,d\theta$ , and $\cos\theta=\sqrt{1-t^2}$ to pass between the two displayed forms. The second kind measures (essentially) arc length of an ellipse:

3

\quad E(\varphi,k)=\int_0^\varphi \sqrt{1-k^2\sin^2\theta}\;d\theta.

Derivation hint: Parametrize an ellipse $x=a\sin\theta,\,y=b\cos\theta$ with $k^2=1-b^2/a^2$ ; the arc-length element $\sqrt{x'^2+y'^2}\,d\theta$ yields $a\sqrt{1-k^2\sin^2\theta}$ . The third kind introduces a parameter $n$ :

4

\quad \Pi(n;\varphi,k)=\int_0^\varphi \frac{d\theta}{(1-n\sin^2\theta)\sqrt{1-k^2\sin^2\theta}}.

Derivation hint: Arises from partial-fraction decomposition of a rational $R$ having a simple pole; the factor $(1-n\sin^2\theta)^{-1}$ carries that pole.

DefinitionComplete Elliptic Integrals

Setting $\varphi=\pi/2$ gives the complete integrals, functions of $k$ alone:

5

\quad K(k)=F\!\left(\tfrac{\pi}{2},k\right)=\int_0^{\pi/2}\frac{d\theta}{\sqrt{1-k^2\sin^2\theta}},\qquad E(k)=\int_0^{\pi/2}\sqrt{1-k^2\sin^2\theta}\;d\theta.

Derivation hint: Quarter-period of the corresponding elliptic function; $K$ plays the role that $\pi/2$ plays for $\sin$ . The complementary modulus is $k'=\sqrt{1-k^2}$ , and $K'(k)=K(k')$ . As $k\to 0$ , $K\to\pi/2$ ; as $k\to 1^-$ , $K\to\infty$ logarithmically.

DefinitionJacobi Elliptic Functions

Invert $F$ : if $u=F(\varphi,k)$ , the amplitude is $\varphi=\operatorname{am}(u,k)$ , and the three primary Jacobi elliptic functions are

6

\quad \operatorname{sn}(u,k)=\sin\varphi,\qquad \operatorname{cn}(u,k)=\cos\varphi,\qquad \operatorname{dn}(u,k)=\sqrt{1-k^2\sin^2\varphi}.

Derivation hint: $\operatorname{sn}$ is defined as the inverse-integral analogue of $\sin$ ; the algebraic relations follow directly from $\sin^2+\cos^2=1$ . They satisfy the Pythagorean-type identities

7

\quad \operatorname{sn}^2 u+\operatorname{cn}^2 u=1,\qquad k^2\operatorname{sn}^2 u+\operatorname{dn}^2 u=1,

Derivation hint: Immediate from [6]: square the first two definitions; use $\operatorname{dn}^2=1-k^2\operatorname{sn}^2$ for the second. with derivatives forming a closed system:

8

\quad \frac{d}{du}\operatorname{sn} u=\operatorname{cn} u\,\operatorname{dn} u,\quad \frac{d}{du}\operatorname{cn} u=-\operatorname{sn} u\,\operatorname{dn} u,\quad \frac{d}{du}\operatorname{dn} u=-k^2\operatorname{sn} u\,\operatorname{cn} u.

Derivation hint: Differentiate $u=\int_0^\varphi d\theta/\sqrt{1-k^2\sin^2\theta}$ to get $d\varphi/du=\sqrt{1-k^2\sin^2\varphi}=\operatorname{dn}u$ , then apply the chain rule to $\sin\varphi,\cos\varphi$ .

TheoremDouble Periodicity

The Jacobi functions are doubly periodic. With $K=K(k)$ and $K'=K(k')$ ,

9

\quad \operatorname{sn}(u+4K)=\operatorname{sn}u,\quad \operatorname{sn}(u+2iK')=\operatorname{sn}u,

so $\operatorname{sn}$ has periods $4K$ and $2iK'$ . More fundamentally, every elliptic function (meromorphic and doubly periodic with period lattice $\Lambda=\mathbb{Z}\omega_1+\mathbb{Z}\omega_2$ , $\omega_2/\omega_1\notin\mathbb{R}$ ) is a rational function of the Weierstrass ℘-function and its derivative. Derivation hint: Real period from completing a full real cycle ( $4$ quarter-periods $K$ ); imaginary period from contour deformation onto the complementary cut governed by $K'$ .

ProofDouble Periodicity from the Integral

We show $\operatorname{sn}$ acquires two periods.

Step 1 (Real period). The real integral $F(\varphi,k)$ increases as $\varphi$ runs over $[0,\pi/2]$ , accumulating $K$ . Over a full real cycle $\varphi:0\to 2\pi$ , $\sin\varphi$ returns to its start after the integral advances by $4K$ (four quarter-periods, by the symmetry $\theta\mapsto\pi-\theta$ etc.). Hence $\operatorname{sn}(u+4K)=\operatorname{sn}u$ .

Step 2 (Imaginary period). Apply the substitution $\theta\to i\psi$ , equivalently analyze the integrand on the path where the radicand passes through the complementary factor $(1-k'^2\sin^2\psi)$ . The total integral over the complementary loop equals $K'=K(k')$ . Analytic continuation of $F$ around the branch points $t=\pm1,\pm1/k$ of the quartic adds $2iK'$ to $u$ while leaving $\operatorname{sn}$ unchanged.

Step 3 (Lattice). Combining, $u$ and $u+m(4K)+n(2iK')$ give the same $\operatorname{sn}$ for all integers $m,n$ . The set $\Lambda$ of such translations is a rank-2 lattice; $\operatorname{sn}$ descends to a meromorphic function on the torus $\mathbb{C}/\Lambda$ .

Step 4 (Why only degree ≥3 works). The Riemann surface of $w=\sqrt{P(x)}$ has genus $\lfloor(\deg P-1)/2\rfloor=1$ for $\deg P\in\{3,4\}$ . Genus 1 is topologically a torus, whose universal cover is $\mathbb{C}$ with a rank-2 deck group — precisely the period lattice. Quadratic $P$ gives genus 0 (a sphere), only one period, recovering trig functions. $\blacksquare$

∎

ProofReduction to Legendre Forms (sketch)

Given $\int R(x,y)\,dx$ with $y^2=P(x)$ quartic:

Step 1. Write $R(x,y)=R_1(x)+R_2(x)\,y^{-1}$ by rationalizing, since $y^2$ is rational in $x$ . The $R_1$ term integrates by elementary means.

Step 2. Expand $R_2(x)/y$ by partial fractions in $x$ ; integrals $\int x^m\,dx/y$ for $m\ge2$ reduce via a recurrence (integration by parts using $d(x^{m-1}y)$ ) down to $m=0,1$ .

Step 3. The surviving irreducible integrals are $\int dx/y$ , $\int x\,dx/y$ , and $\int dx/((x-c)y)$ — which, after a Möbius substitution sending the roots of $P$ to $\pm1,\pm1/k$ , become exactly $F$ , $E$ , and $\Pi$ . $\blacksquare$

∎

CG

Carl Gustav Jacob Jacobi

1804–1851

Jacobi, alongside the tragically short-lived Niels Henrik Abel, founded the theory of elliptic functions in the late 1820s. His 1829 treatise Fundamenta nova theoriae functionum ellipticarum introduced the sn, cn, dn notation, theta functions, and the transformation theory connecting different moduli. Where Legendre had spent forty years cataloguing the integrals, Jacobi's decisive insight — shared independently with Abel — was to study their inverses, revealing double periodicity. His work seeded the theory of automorphic functions, modular forms, and ultimately the arithmetic of elliptic curves.

🧠 Intuition

Think of $\sin$ as "the inverse of the simplest arc-length integral on a circle." The circle has one way to wrap around, so $\sin$ has one period $2\pi$ . Now imagine the natural geometric object behind a quartic radical: a torus (donut). A torus has two independent loops — around the hole and through the hole. A function natural to the torus must therefore repeat in two independent directions, giving two periods. Elliptic functions are simply "trigonometry on a donut." The modulus $k$ controls the donut's shape: $k\to0$ pinches it toward a circle (recovering trig, $\operatorname{sn}\to\sin$ ), while $k\to1$ stretches one period to infinity (recovering hyperbolic/exponential behavior, $\operatorname{sn}\to\tanh$ ).

Figure

The period lattice of an elliptic function: sn(u) repeats over a rectangular grid with real period 4K and imaginary period 2iK′

ExamplePeriod of a Simple Pendulum (large amplitude)

Given: A pendulum of length $L=1.00$ m, $g=9.81$ m/s², released from rest at amplitude $\theta_0=90^\circ=\pi/2$ .

Setup. The exact period is

10

\quad T=4\sqrt{\frac{L}{g}}\,K(k),\qquad k=\sin\!\frac{\theta_0}{2}.

Derivation hint: Energy conservation gives $\dot\theta=\sqrt{(2g/L)(\cos\theta-\cos\theta_0)}$ ; substitute $\sin(\theta/2)=k\sin\phi$ to obtain $K(k)$ .

Steps. 1. $k=\sin(45^\circ)=0.7071$ . 2. From tables/series, $K(0.7071)=1.8541$ . 3. $\sqrt{L/g}=\sqrt{1.00/9.81}=0.3193$ s. 4. $T=4(0.3193)(1.8541)=2.368$ s. Compare the small-angle approximation $T_0=2\pi\sqrt{L/g}=2.006$ s — the exact period is 18% longer.

\boxed{T \approx 2.37\ \text{s}}

ExampleArc Length of an Ellipse

Given: An ellipse with semi-axes $a=5$ , $b=3$ (dimensionless). Find its full perimeter.

Setup. The perimeter is $P=4a\,E(k)$ with $k=\sqrt{1-b^2/a^2}$ (eccentricity).

11

\quad P=4a\int_0^{\pi/2}\sqrt{1-k^2\sin^2\theta}\,d\theta=4a\,E(k).

Derivation hint: From [3], integrate the arc-length element over one quadrant and multiply by 4.

Steps. 1. $k^2=1-9/25=0.64$ , so $k=0.80$ . 2. $E(0.80)=1.2764$ (from tables/series). 3. $P=4(5)(1.2764)=25.53$ . Check: Ramanujan's approximation $P\approx\pi[3(a+b)-\sqrt{(3a+b)(a+3b)}]=25.527$ — excellent agreement.

\boxed{P \approx 25.53\ \text{(length units)}}

⚠️ Common mistake

"Elliptic integrals integrate over ellipses." Only the second kind $E$ relates directly to ellipse arc length, and that's a historical accident giving the whole subject its name. Elliptic integrals and functions appear in countless contexts having nothing to do with ellipses (pendulums, vortices, lemniscates, modular forms). The defining feature is the quartic/cubic under the radical, not any ellipse.

⚠️ Common mistake

"The modulus $k$ and the parameter $m=k^2$ are interchangeable notation." They are not. Many references (and software like Mathematica) use $m=k^2$ as the argument: $K(m)$ vs $K(k)$ . Always check whether a source writes $E(k)$ or $E(m)$ — confusing them produces silently wrong numbers. The same caution applies to the third-kind sign convention for $n$ .

Counterexample

A non-existent elementary antiderivative. One might hope $\int_0^{\pi/2} d\theta/\sqrt{1-\tfrac12\sin^2\theta}$ has a closed elementary form. By Liouville's theorem on integration in finite terms, it provably does not: the integrand's associated Riemann surface has genus 1, and Liouville/Risch theory shows no elementary antiderivative exists. The integral equals $K(1/\sqrt2)=1.8541\ldots$ , an essentially new transcendental constant (related to the lemniscate constant $\varpi$ ). This is the precise sense in which elliptic integrals are "not elementary."

connection

Elliptic integrals sit at a crossroads. Letting $\deg P=2$ collapses them to Trigonometric Functions and Inverse Trigonometric Functions — elliptic theory is their genus-1 generalization. The inverse functions connect to Complex Analysis through meromorphic doubly-periodic functions on a Riemann Surface (genus-1 torus). The Weierstrass ℘-function parametrizes Elliptic Curves $y^2=4x^3-g_2x-g_3$ , linking analysis to Number Theory and the arithmetic of rational points. The arithmetic–geometric mean gives a quadratically convergent algorithm for $K(k)$ , tying the subject to Numerical Analysis. Finally, theta functions and the modular parameter $\tau=iK'/K$ open onto Modular Forms.

🔬 Where this appears

⏳ History

- 1655–1694 — John Wallis and Jacob Bernoulli encounter elliptic arc-length integrals; Bernoulli studies the elastica and lemniscate, whose arc length

\int dt/\sqrt{1-t^4}

resists elementary evaluation.

- 1718 — Giulio Fagnano discovers the duplication formula for lemniscatic arcs, the first hint of an addition structure.

- 1751 — Euler, prompted by Fagnano's work, proves the general addition theorem for elliptic integrals — a landmark result.

- 1786–1827 — Adrien-Marie Legendre devotes decades to systematizing elliptic integrals, classifying them into three kinds and producing extensive tables (Traité des fonctions elliptiques).

- 1827 — Niels Henrik Abel, studying the inverse function, discovers double periodicity (Recherches sur les fonctions elliptiques); Abel's theorem generalizes far beyond.

- 1829 — Jacobi publishes Fundamenta nova, introducing sn/cn/dn, theta functions, and transformation theory; he and Abel are recognized as co-founders.

- 1847–1856 — Eisenstein and Weierstrass develop the ℘-function approach, giving the cleanest modern foundation via the lattice

\Lambda

.

- 1862 — Weierstrass's Berlin lectures crystallize the theory of

\wp,\wp'

and the differential equation

\wp'^2=4\wp^3-g_2\wp-g_3

.

- 20th–21st c. — Elliptic curves become central to number theory (Mordell, Weil, Wiles's proof of Fermat's Last Theorem via modularity) and to cryptography.

open-problem

Higher genus and hyperelliptic generalizations. Replacing the quartic $P$ by a polynomial of degree $\ge5$ yields hyperelliptic integrals whose inverses are Abelian functions of several complex variables, living on Jacobians of higher-genus curves. Unlike the genus-1 case, no single function plays the universal role ℘ does, and explicit, computationally tractable analogues remain an active research frontier (the Schottky problem of characterizing Jacobians among all abelian varieties is deep and only partially resolved). Even for elliptic curves, fine arithmetic questions — the Birch–Swinnerton-Dyer conjecture relating $L$ -functions to ranks of rational points — remain among the Clay Millennium Problems, showing that the "simple" genus-1 case is far from exhausted.

Elliptic Integrals and Elliptic Functions

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also