Elliptic Functions — Weierstrass p-function

⚡ Key Takeaways

The Weierstrass ℘-function is the canonical doubly-periodic meromorphic function on the complex plane, with a double pole at each point of a lattice $\Lambda = \mathbb{Z}\omega_1 + \mathbb{Z}\omega_2$ . It is the fundamental building block of elliptic function theory: every elliptic function with period lattice $\Lambda$ is a rational function of $\wp$ and its derivative $\wp'$ . Used in algebraic geometry (elliptic curves), number theory, integrable systems, and mathematical physics. Introduced by Karl Weierstrass in lectures of the 1860s–1880s.

💡 The Core Insight

A non-constant doubly-periodic meromorphic function cannot be holomorphic everywhere (it would be bounded, hence constant by Liouville). The minimal-pole option, a single double pole per fundamental cell, is realized by $\wp$ . Because $\wp$ is even and $\wp'$ is odd, the pair $(\wp, \wp')$ parametrizes a plane cubic, and the addition of points on $\mathbb{C}/\Lambda$ becomes the chord-and-tangent group law on that cubic. Periodicity + minimal poles = elliptic curve.

DefinitionLattice and Elliptic Functions

A lattice is $\Lambda = \mathbb{Z}\omega_1 + \mathbb{Z}\omega_2$ where $\omega_1, \omega_2 \in \mathbb{C}$ are $\mathbb{R}$ -linearly independent (equivalently $\tau = \omega_2/\omega_1 \notin \mathbb{R}$ ; WLOG $\operatorname{Im}\tau > 0$ ). A meromorphic function $f:\mathbb{C}\to\mathbb{C}\cup\{\infty\}$ is elliptic (doubly periodic) with respect to $\Lambda$ if

1

\quad f(z+\omega) = f(z) \qquad \forall\, \omega \in \Lambda,\ \forall z.

Derivation hint: it suffices to require invariance under the two generators $\omega_1, \omega_2$ , which then extends to all integer combinations. A fundamental parallelogram (period cell) is $P = \{a\omega_1 + b\omega_2 : a,b \in [0,1)\}$ ; it tiles $\mathbb{C}$ under translation by $\Lambda$ .

DefinitionThe Weierstrass ℘-function

For a lattice $\Lambda$ , define

2

\quad \wp(z;\Lambda) = \frac{1}{z^2} + \sum_{\substack{\omega \in \Lambda \\ \omega \neq 0}} \left( \frac{1}{(z-\omega)^2} - \frac{1}{\omega^2} \right).

Derivation hint: the bare sum $\sum (z-\omega)^{-2}$ diverges like $\sum |\omega|^{-2}$ ; subtracting $\omega^{-2}$ makes each term $O(|\omega|^{-3})$ for fixed $z$ , restoring convergence. Here $z \in \mathbb{C}\setminus\Lambda$ . The function $\wp$ is meromorphic on $\mathbb{C}$ with a double pole (no residue) at every lattice point and is holomorphic elsewhere.

TheoremFundamental Properties of ℘

Let $\wp = \wp(\,\cdot\,;\Lambda)$ . Then: 1. $\wp$ is even: $\wp(-z) = \wp(z)$ . 2. $\wp$ is $\Lambda$ -periodic and meromorphic, with order 2 (two poles per cell, counted with multiplicity). 3. Its derivative is

3

\quad \wp'(z) = -2 \sum_{\omega \in \Lambda} \frac{1}{(z-\omega)^3},

Derivation hint: differentiate [2] term by term; the now-cubic terms converge absolutely so no counterterms are needed, and the $\omega = 0$ term recombines naturally. which is odd and $\Lambda$ -periodic of order 3. 4. $\wp$ satisfies the algebraic differential equation

4

\quad \big(\wp'(z)\big)^2 = 4\wp(z)^3 - g_2 \wp(z) - g_3,

Derivation hint: form $\wp'^2 - 4\wp^3 + g_2\wp + g_3$ ; choose $g_2, g_3$ so the principal parts at $z=0$ cancel; the result is elliptic, holomorphic, hence constant $=0$ by Liouville. where the Eisenstein invariants are

5

\quad g_2 = 60 \sum_{\omega\neq 0} \frac{1}{\omega^4} = 60\, G_4, \qquad g_3 = 140 \sum_{\omega\neq 0} \frac{1}{\omega^6} = 140\, G_6.

Derivation hint: expand $\wp$ in a Laurent series about $z=0$ ; the coefficients are the Eisenstein series $G_{2k}=\sum_{\omega\neq0}\omega^{-2k}$ , and matching cubic and linear poles fixes $g_2, g_3$ .

ProofConvergence, periodicity, and the ODE

Step 1 — Convergence of [2]. For $|z| \le R$ and $|\omega| > 2R$ ,

7

\quad \left| \frac{1}{(z-\omega)^2} - \frac{1}{\omega^2} \right| = \left| \frac{z(2\omega - z)}{\omega^2 (z-\omega)^2} \right| \le \frac{C\,|z|}{|\omega|^3}.

Derivation hint: combine fractions over a common denominator and bound $|2\omega - z| \le 3|\omega|$ , $|z-\omega| \ge |\omega|/2$ . The lattice sum $\sum_{\omega\neq 0} |\omega|^{-s}$ converges iff $s > 2$ (comparison with $\int r^{-s}\, r\, dr\, d\theta$ ). Hence the series converges absolutely and uniformly on compact sets avoiding $\Lambda$ , defining a meromorphic function with double poles exactly on $\Lambda$ .

Step 2 — Evenness. Replacing $z \to -z$ and reindexing $\omega \to -\omega$ (a bijection of $\Lambda$ ) leaves [2] invariant, so $\wp(-z) = \wp(z)$ .

Step 3 — Periodicity. From [3], $\wp'$ is manifestly $\Lambda$ -periodic since $\omega \mapsto \omega - \omega_j$ permutes $\Lambda$ . Thus $\wp(z+\omega_j) - \wp(z) = c_j$ is constant. Set $z = -\omega_j/2$ : by evenness $\wp(\omega_j/2) = \wp(-\omega_j/2)$ , giving $c_j = 0$ . So $\wp$ is $\Lambda$ -periodic.

Step 4 — The differential equation. Using [6],

8

\quad \wp(z)^3 = z^{-6} + 9G_4 z^{-2} + 15 G_6 + O(z^2), \qquad \wp'(z)^2 = 4z^{-6} - 24 G_4 z^{-2} - 80 G_6 + O(z^2).

Derivation hint: cube [6] and square its term-by-term derivative $\wp'(z) = -2z^{-3} + 6G_4 z + 20 G_6 z^3 + \cdots$ , keeping terms up to constants. Form $F(z) = \wp'^2 - 4\wp^3 + 60 G_4 \wp + 140 G_6$ . By [8] all negative powers and the constant term cancel, so $F$ is holomorphic at $z=0$ with $F(0)=0$ . Being elliptic with no poles, $F$ is bounded, hence constant by Liouville, hence $F \equiv 0$ . This is [4] with $g_2 = 60G_4$ , $g_3 = 140 G_6$ . $\blacksquare$

∎

TheoremLiouville's Theorems for Elliptic Functions

Let $f$ be elliptic with period cell $P$ : 1. If $f$ is holomorphic on all of $\mathbb{C}$ , then $f$ is constant. 2. $\sum_{\text{poles in } P} \operatorname{Res} f = 0$ . 3. The number of zeros equals the number of poles in $P$ (the order of $f \ge 2$ if non-constant). 4. $\sum (\text{zeros}) \equiv \sum (\text{poles}) \pmod{\Lambda}$ .

ProofSketch of Liouville's theorems

(1) A holomorphic elliptic function is continuous on the compact closure $\bar P$ , hence bounded, hence bounded on $\mathbb{C}$ by periodicity; Liouville's theorem (entire + bounded ⇒ constant) finishes it.

(2) Integrate $f$ over the boundary $\partial P$ . Opposite sides cancel by periodicity, so $\oint_{\partial P} f\, dz = 0 = 2\pi i \sum \operatorname{Res}$ . (Shift $P$ slightly so no pole sits on $\partial P$ .)

(3) Apply (2) to $f'/f$ , whose residues count zeros minus poles.

(4) Apply the same boundary argument to $z\, f'(z)/f(z)$ ; the non-cancelling boundary terms produce a lattice vector. $\blacksquare$

∎

mathematician · Karl Weierstrass | 1815–1897

Weierstrass, the "father of modern analysis," rebuilt the theory of elliptic functions from his Berlin lectures (1860s onward) on a rigorous, systematic footing, replacing Jacobi's theta-quotient formalism with the single privileged function $\wp$ and its sibling sigma and zeta functions. His insistence on $\varepsilon$ – $\delta$ rigor, uniform convergence, and power-series foundations shaped the entire subsequent development of complex analysis. Though he published little of this material himself, it spread through students like Sofia Kovalevskaya, Hermann Schwarz, and Adolf Hurwitz, and through the lecture notes that became canonical.

🧠 Intuition

Think of $\wp$ as the "electrostatic potential" of a doubly-periodic array of identical charges. Each lattice point hosts a singularity ( $1/(z-\omega)^2$ ); periodicity demands an infinite array. A naive infinite sum of equal contributions diverges, so we subtract the constant background $1/\omega^2$ — exactly like renormalizing a periodic potential against a uniform charge sheet. The result is the unique even, order-2 elliptic function regular at the origin up to its prescribed double pole. The map $z \mapsto (\wp(z), \wp'(z))$ wraps the torus $\mathbb{C}/\Lambda$ onto the cubic curve [4]. The torus has genus 1; so does the curve. The two "missing" points ( $z=0$ , the pole) both map to the single point at infinity of the projective cubic — the wrap is a bijection of $\mathbb{C}/\Lambda$ onto the curve. Addition of complex numbers mod $\Lambda$ becomes the group law on the curve, and that is the deep payoff: a transcendental uniformization of an algebraic group.

Figure

The lattice Λ, fundamental cell, half-periods, and the ℘-map to the cubic

```svg

```

ExampleThe lemniscatic (square) lattice: half-period values

Given: The lattice $\Lambda = \mathbb{Z} + \mathbb{Z}i$ (square lattice, $\tau = i$ ). By symmetry under $z \mapsto iz$ (which permutes $\Lambda$ ), the Eisenstein series $G_6$ over a square lattice vanishes: rotating by $i$ multiplies $G_6$ by $i^{-6} = -1$ , forcing $G_6 = 0$ , hence $g_3 = 0$ .

Goal: Find $\wp$ at the half-periods $e_1 = \wp(1/2)$ , $e_2 = \wp((1+i)/2)$ , $e_3 = \wp(i/2)$ .

Steps: 1. The cubic [4] factors as $\wp'^2 = 4(\wp - e_1)(\wp - e_2)(\wp - e_3)$ where $e_i$ are the values at the three half-periods (these are exactly the points where $\wp' = 0$ , since $\wp'$ is odd and periodic). 2. With $g_3 = 0$ , the cubic is $4x^3 - g_2 x = 4x(x^2 - g_2/4)$ . Its roots are

9

\quad e_2 = 0, \qquad e_1 = +\tfrac12\sqrt{g_2}, \qquad e_3 = -\tfrac12\sqrt{g_2}.

Derivation hint: solve $4e^3 - g_2 e = 0$ ; the central root is $0$ , the others are symmetric. 3. For $\Lambda = \mathbb{Z}+\mathbb{Z}i$ , numerically $g_2 = 60 G_4 \approx 189.07$ , so $\sqrt{g_2}/2 \approx 6.875$ .

Answer:

\boxed{\,e_2 = 0,\quad e_1 \approx +6.875,\quad e_3 \approx -6.875\,}

(dimensionless; $e_1 + e_2 + e_3 = 0$ always, since the $x^2$ coefficient of $4x^3 - g_2 x - g_3$ vanishes).

ExampleAddition formula: computing ℘(2u)

Given: A point $u \in \mathbb{C}/\Lambda$ with known $\wp(u) = p$ and $\wp'(u) = q$ satisfying $q^2 = 4p^3 - g_2 p - g_3$ . We want the duplication value $\wp(2u)$ .

Steps: 1. The addition theorem states

10

\quad \wp(u+v) = \frac14\left(\frac{\wp'(u)-\wp'(v)}{\wp(u)-\wp(v)}\right)^2 - \wp(u) - \wp(v).

Derivation hint: $\wp(u+v) + \wp(u) + \wp(v)$ equals the square of the slope of the line through $(\wp(u),\wp'(u))$ and $(\wp(v),\wp'(v))$ , divided by 4 — the chord-and-tangent group law on the cubic. 2. Taking the limit $v \to u$ , the slope becomes the tangent slope $\wp''(u)/\wp'(u)$ , giving the duplication formula

11

\quad \wp(2u) = \frac14\left(\frac{\wp''(u)}{\wp'(u)}\right)^2 - 2\wp(u).

Derivation hint: differentiate [4] to get $2\wp'\wp'' = 12\wp^2 - g_2$ , so $\wp'' = 6\wp^2 - g_2/2$ ; substitute. 3. Therefore, in terms of $p,q$ with $\wp'' = 6p^2 - g_2/2$ :

12

\quad \wp(2u) = \frac{(6p^2 - g_2/2)^2}{4q^2} - 2p = \frac{(12p^2 - g_2)^2}{16(4p^3 - g_2 p - g_3)} - 2p.

Derivation hint: substitute $\wp'' = 6p^2-g_2/2$ and $q^2 = 4p^3-g_2 p - g_3$ into [11] and simplify.

Numerical instance: Take $g_2 = 4$ , $g_3 = 0$ , and $p = \wp(u) = 1$ . Then $q^2 = 4(1) - 4(1) - 0 = 0$ — this is a half-period (where $\wp'=0$ ), so we use [11] differently; instead take $p = 2$ : $q^2 = 4\cdot8 - 4\cdot2 = 24$ . Then $12p^2 - g_2 = 48 - 4 = 44$ , and

\wp(2u) = \frac{44^2}{16\cdot 24} - 4 = \frac{1936}{384} - 4 = 5.0417 - 4.

Answer:

\boxed{\,\wp(2u) \approx 1.0417\,}

(dimensionless), illustrating how doubling on the torus is a rational map on the cubic.

⚠️ Common mistake

$\wp$ does not have a simple pole. Its poles are double poles with zero residue. A non-constant elliptic function of order 1 (a single simple pole per cell) is impossible: by Liouville's second theorem the residues in a cell sum to zero, so a single simple pole would need residue 0, contradicting simplicity. The minimal non-constant elliptic functions therefore have order 2 — either one double pole (like $\wp$ ) or two distinct simple poles (like Jacobi's functions in their own normalization).

⚠️ Common mistake

Eisenstein series $G_2$ is special. The series $\sum_{\omega\neq 0}\omega^{-2}$ is only conditionally convergent and is not modular of weight 2; it transforms with an anomalous additive term. This is why $\wp$ uses the absolutely convergent counterterm construction [2] rather than naively $\sum (z-\omega)^{-2}$ , and why $g_2, g_3$ start at weights 4 and 6. The would-be " $G_2$ " reappears as the Weierstrass $\eta$ quasi-period and in the non-holomorphic $E_2^*$ .

Counterexample

One might guess that any even meromorphic doubly-periodic function with one double pole equals $\wp$ up to scaling. False without normalizing the principal part. The function $\wp(z) + c$ for any constant $c$ is also even, $\Lambda$ -periodic, with a single double pole of leading coefficient $1$ — so it is not unique among such functions; uniqueness requires fixing the full principal part $z^{-2}$ at the lattice point and the value (or the absence of a constant term in the Laurent expansion, as in [6]). Concretely, $\wp(z) + 7$ and $\wp(z)$ satisfy different ODEs ([4] with shifted $g_3$ is not even valid — shifting $\wp$ does not preserve the cubic form), so "looks like $\wp$ " is not enough.

connection

The $\wp$ -function sits at the crossroads of several theories:

Elliptic Curves: The map $z \mapsto [\wp(z):\wp'(z):1]$ is an analytic isomorphism $\mathbb{C}/\Lambda \xrightarrow{\sim} E(\mathbb{C})$ where $E: y^2 = 4x^3 - g_2 x - g_3$ . Group addition on $\mathbb{C}/\Lambda$ matches the chord–tangent law.
Modular Forms: $g_2$ and $g_3$ are modular forms of weights 4 and 6; the discriminant $\Delta = g_2^3 - 27 g_3^2$ is a weight-12 cusp form, and the $j$ -invariant $j = 1728\, g_2^3/\Delta$ classifies lattices up to homothety.
Theta Functions: $\wp$ can be written via Jacobi theta quotients; the Weierstrass $\sigma$ -function is essentially $\theta_1$ , with $\wp = -(\log\sigma)''$ .
Jacobi Elliptic Functions: $\operatorname{sn},\operatorname{cn},\operatorname{dn}$ are related to $\wp$ by Möbius changes of variable; both describe the same tori in different normalizations.
Riemann Surfaces: $\mathbb{C}/\Lambda$ is the universal example of a compact genus-1 Riemann surface, and $\wp$ provides its embedding into $\mathbb{P}^2$ .

🔬 Where this appears

⏳ History

- 1750s–1786 — Euler & Lagrange: Addition theorems for elliptic integrals discovered; Euler's addition formula for

\int dx/\sqrt{\text{quartic}}

foreshadows the group law.

- 1820s — Abel & Jacobi: Independently invert elliptic integrals to obtain doubly-periodic functions; Jacobi's Fundamenta Nova (1829) introduces theta functions and

\operatorname{sn},\operatorname{cn},\operatorname{dn}

.

- 1844–1847 — Eisenstein & Liouville: Eisenstein studies the lattice series

\sum 1/(z-\omega)^n

; Liouville proves the foundational theorems on elliptic functions (boundedness ⇒ constant, residue sum zero).

- 1850s — Hermite, Briot & Bouquet: Systematize the field-theoretic structure of elliptic functions.

- 1862–1863 — Weierstrass: Lectures in Berlin introduce

\wp

,

\sigma

,

\zeta

, the invariants

g_2, g_3

, and the canonical ODE; rigor and the single privileged function replace theta quotients.

- 1882 — Schwarz: Publishes lecture notes (Formeln und Lehrsätze) disseminating Weierstrass's theory widely.

- 1893 — Halphen: Traité des fonctions elliptiques gives a comprehensive synthesis with applications to mechanics.

- 1920s–1930s — Hecke, Hurwitz, Klein: Modular interpretation of

g_2, g_3, \Delta, j

; the bridge to modular forms is cemented.

- 1985–present — ECC era: Koblitz and Miller propose elliptic-curve cryptography; the

\wp

-uniformized group law becomes computationally central.

open-problem

While $\wp$ itself is fully understood, the arithmetic it uncovers fuels open problems:

Birch–Swinnerton-Dyer Conjecture: Relates the rank of the Mordell–Weil group $E(\mathbb{Q})$ — points on the Weierstrass cubic — to the order of vanishing of the $L$ -function $L(E,s)$ at $s=1$ . A Clay Millennium Problem.
Lehmer-type questions on $\Delta$ and $\eta$ : Non-vanishing and size of Fourier coefficients of the discriminant (Ramanujan's $\tau$ -function, $\Delta = q\prod(1-q^n)^{24}$ ) — whether $\tau(n)\neq 0$ for all $n$ remains open.
Periods and transcendence: $g_2, g_3$ , the periods $\omega_1,\omega_2$ , and quasi-periods $\eta_1,\eta_2$ obey the Legendre relation $\eta_1\omega_2 - \eta_2\omega_1 = 2\pi i$ , but the transcendence theory (Chudnovsky's results, conjectures of Grothendieck on periods) is far from complete.
Effective uniformization in higher genus: Generalizing $\wp$ to abelian varieties (Klein/Kleinian $\wp$ -functions, hyperelliptic sigma functions) is an active area with incomplete algorithmic understanding.

Elliptic Functions — Weierstrass p-function

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also