06.01.31 · riemann-surfaces / complex-analysis

Jacobi theta functions and the triple product

shipped3 tiersLean: none

Anchor (Master): Mumford *Tata Lectures on Theta I* (Birkhauser 1983) Ch. I; Jacobi 1829 *Fundamenta Nova Theoriae Functionum Ellipticarum* (Konigsberg); Whittaker-Watson Ch. 21; Chandrasekharan *Elliptic Functions* (Springer 1985) Ch. V

Intuition Beginner

A periodic function repeats when you slide its input by a fixed step: $sin (x)$ comes back to itself every $2 π$ . The Jacobi theta functions are a family of four functions built to be the master raw material out of which doubly-periodic functions on a lattice are assembled. Each one is an infinite sum of waves whose sizes shrink very fast, so the sum converges and defines a smooth complex function.

The shrinking is controlled by a single number $q$ , called the nome, with $∣ q ∣ < 1$ . Think of $q$ as a dial. When $q$ is close to $0$ , only the first wave matters and the function is nearly constant. As $q$ grows toward the edge $∣ q ∣ = 1$ , more and more waves contribute and the function develops sharp, repeating features.

The four functions are close cousins: $θ_{3}$ is the plain version, and $θ_{1}, θ_{2}, θ_{4}$ are obtained by small shifts and sign changes. One of them, $θ_{1}$ , is special because it has a single zero at the centre, which makes it the right building block for functions that need to vanish at prescribed places.

Why do these four functions exist as a named family? Because they package three deep facts at once: a product formula that factors each sum into a clean infinite product, a set of transformation rules that say how they behave when the dial $q$ is changed in a structured way, and a web of polynomial identities relating their central values. Those three facts underlie the heat equation, elliptic functions, and modular forms.

Visual Beginner

Picture the complex plane and a sum of cosine-like waves stacked on top of one another. The first wave is tall and slow; each later wave is shorter and faster, scaled down by a higher power of the nome $q$ . Adding them gives a single bumpy curve that repeats horizontally with period $π$ and, when you shift in the slanted lattice direction, repeats up to a predictable stretching factor.

The picture makes one thing vivid: the four theta functions differ only by where their peaks and zeros sit. Sliding the whole curve by a quarter-period or a half-period turns one theta function into another, which is the visual shadow of the algebraic identities relating them.

Worked example Beginner

Take $θ_{3}$ at the centre $z = 0$ , where it becomes a pure power series in the nome:

θ_{3} (0 ∣ τ) = 1 + 2 q + 2 q^{4} + 2 q^{9} + 2 q^{16} + \dots,

with one term $2 q^{n^{2}}$ for each positive whole number $n$ , plus the leading $1$ . The exponents are the perfect squares $1, 4, 9, 16$ .

Step 1. Pick a small nome, say $q = 0.1$ . Then $q = 0.1$ , $q^{4} = 0.0001$ , $q^{9}$ is a billionth, and everything after is negligible.

Step 2. Add the surviving terms: $1 + 2 (0.1) + 2 (0.0001) = 1 + 0.2 + 0.0002 = 1.2002$ . So $θ_{3} (0 ∣ τ) \approx 1.2002$ for this nome.

Step 3. Notice the structure: the function counts squares. The coefficient of $q^{n}$ in $θ_{3} (0)$ records how many ways $n$ is a sum of squares, which is why theta functions tie complex analysis to counting problems in number theory.

What this tells us: even at a single point, a theta function is an organised bookkeeping device. The fast shrinkage of $q^{n^{2}}$ makes the value easy to compute, and the exponents being squares is the seed of the deep modular transformation rules covered in the later tiers.

Check your understanding Beginner

Formal definition Intermediate+

Fix $τ$ in the upper half-plane $H = {τ : Im τ > 0}$ and set the nome $q = e^{π i τ}$ , so $∣ q ∣ < 1$ . The four Jacobi theta functions of $z \in C$ are

θ_{3} (z ∣ τ) = n = - \infty \sum \infty q^{n^{2}} e^{2 in z}, θ_{4} (z ∣ τ) = n = - \infty \sum \infty (- 1)^{n} q^{n^{2}} e^{2 in z},

θ_{2} (z ∣ τ) = n = - \infty \sum \infty q^{(n + \frac{1}{2})^{2}} e^{(2 n + 1) i z}, θ_{1} (z ∣ τ) = - i n = - \infty \sum \infty (- 1)^{n} q^{(n + \frac{1}{2})^{2}} e^{(2 n + 1) i z} .

Each series converges absolutely and uniformly on compact subsets of $C \times H$ , since $∣ q^{n^{2}} ∣ = e^{- π n^{2} Im τ}$ decays super-exponentially in $n$ ; hence each $θ_{j}$ is holomorphic in both variables. The function $θ_{1}$ is odd in $z$ and the other three are even; $θ_{1}$ has a simple zero at $z = 0$ , while $θ_{2}, θ_{3}, θ_{4}$ are non-zero there.

The quasi-periodicity under the two lattice steps reads, for $θ_{3}$ ,

θ_{3} (z + π ∣ τ) = θ_{3} (z ∣ τ), θ_{3} (z + π τ ∣ τ) = q^{- 1} e^{- 2 i z} θ_{3} (z ∣ τ),

with parallel multipliers for the others (the $π τ$ -shift produces the factor $q^{- 1} e^{- 2 i z}$ up to a sign that distinguishes the four). ^{[Whittaker-Watson Ch. 21]}

The theta-nullwerte are the central values $θ_{2} (0 ∣ τ)$ , $θ_{3} (0 ∣ τ)$ , $θ_{4} (0 ∣ τ)$ ; the fourth, $θ_{1} (0 ∣ τ)$ , vanishes, and the relevant invariant is the derivative $θ_{1}^{'} (0 ∣ τ)$ .

Counterexamples to common slips

The horizontal period is $π$ , not $2 π$ . Each term carries $e^{2 in z}$ , so shifting $z$ by $π$ multiplies $e^{2 in z}$ by $e^{2 π in} = 1$ . Using $2 π$ as the period double-counts the lattice and produces the wrong fundamental domain.
$θ_{1}$ is not even. The $(- 1)^{n}$ weight combined with the half-integer frequencies makes $θ_{1} (- z ∣ τ) = - θ_{1} (z ∣ τ)$ . Treating $θ_{1}$ as even erases its defining simple zero at the origin and breaks every elliptic-function construction that uses it.
The $π τ$ -shift is a multiplier, not a true period. The function is only quasi-periodic in the lattice direction: $θ_{3} (z + π τ) = q^{- 1} e^{- 2 i z} θ_{3} (z)$ acquires a non-constant exponential factor. The theta functions are sections of a line bundle on the torus, not functions on it; ratios of thetas with matching multipliers descend to the torus.

Key theorem with proof Intermediate+

Theorem (Modular transformation of $θ_{3}$ via Poisson summation). For $τ \in H$ ,

θ_{3} (0 - \frac{1}{τ}) = - i τ θ_{3} (0 ∣ τ),

where $\cdot$ is the branch positive on the positive real axis. Together with the period $θ_{3} (0 ∣ τ + 1) = θ_{2} (0 ∣ τ)$ -type relabelling, the two substitutions $τ \mapsto τ + 1$ and $τ \mapsto - 1/ τ$ generate the modular action on the theta-nullwerte, permuting $θ_{2} \leftrightarrow θ_{4}$ under $τ \mapsto - 1/ τ$ .

Proof. Write the central value as a Gaussian sum. With $q = e^{π i τ}$ ,

θ_{3} (0 ∣ τ) = n = - \infty \sum \infty e^{π i τ n^{2}} = n = - \infty \sum \infty f (n), f (x) = e^{π i τ x^{2}} .

For $Im τ > 0$ the function $f$ is a Schwartz function (a Gaussian with negative real part in the exponent after the rotation), so the Poisson summation formula applies:

n \sum f (n) = m \sum \hat{f} (m), \hat{f} (m) = \int_{- \infty}^{\infty} e^{π i τ x^{2}} e^{- 2 π im x} d x .

Complete the square in the exponent: $π i τ x^{2} - 2 π im x = π i τ (x - m / τ)^{2} - π i m^{2} / τ$ . The Gaussian integral evaluates (rotating the contour, valid since $Im τ > 0$ ) to

\hat{f} (m) = \frac{1}{- i τ} e^{- π i m^{2} / τ} .

Substituting back,

θ_{3} (0 ∣ τ) = \frac{1}{- i τ} m \sum e^{- π i m^{2} / τ} = \frac{1}{- i τ} θ_{3} (0 - \frac{1}{τ}),

since replacing $τ$ by $- 1/ τ$ in the defining series sends $e^{π i τ m^{2}}$ to $e^{- π i m^{2} / τ}$ . Rearranging gives $θ_{3} (0 ∣ - 1/ τ) = - i τ θ_{3} (0 ∣ τ)$ .

The same Poisson computation applied to the shifted lattices (half-integer offsets in $n$ or alternating signs) yields the companion rules; the half-integer offset that defines $θ_{2}$ is exchanged with the alternating sign that defines $θ_{4}$ under $τ \mapsto - 1/ τ$ , which is the permutation $θ_{2} \leftrightarrow θ_{4}$ . Under $τ \mapsto τ + 1$ the nome picks up $q \mapsto e^{π i} q = - q$ , fixing $θ_{4}$ and exchanging $θ_{2}$ with $θ_{3}$ up to a root of unity. $□$

Bridge. This transformation law builds toward the entire theory of modular forms 21.04.01 and appears again in the modularity proof for the elliptic genus 03.06.23, where $θ_{1}$ 's rule is the load-bearing input. The foundational reason the rule holds is self-duality of the Gaussian under the Fourier transform: Poisson summation converts a sum over a lattice into a sum over its dual, and the dual of $Z$ rescaled by $τ$ is $Z$ rescaled by $- 1/ τ$ . This is exactly the analytic substrate that the rigidity theorem 03.06.24 invokes without deriving, and the central insight is that modular invariance of physical partition functions is the visible face of this lattice duality. Putting these together, the theta multiplier $- i τ$ is the half-integral weight that makes $η^{24}$ and $Δ$ honest modular forms.

Exercises Intermediate+

Exercise 5 (medium, symbolic).

In the frequency convention used here ( $e^{2 in z}$ in each term), the theta functions satisfy a heat equation $\partial_{τ} θ = c \partial_{z}^{2} θ$ for a fixed constant $c$ . Determine $c$ by matching the term-by-term derivatives for $θ_{3}$ .

Hint

Differentiate $q^{n^{2}} e^{2 in z} = e^{π i τ n^{2}} e^{2 in z}$ once in $τ$ and twice in $z$ , then take the ratio of the two results.

Answer

$c = - \frac{π i}{4}$ , equivalently $4 i \partial_{τ} θ = π \partial_{z}^{2} θ$ . For the general term $u_{n} = e^{π i τ n^{2}} e^{2 in z}$ : $\partial_{τ} u_{n} = π i n^{2} u_{n}$ and $\partial_{z}^{2} u_{n} = (2 in)^{2} u_{n} = - 4 n^{2} u_{n}$ . The ratio is $\partial_{τ} u_{n} / \partial_{z}^{2} u_{n} = π i n^{2} / (- 4 n^{2}) = - π i /4$ , independent of $n$ . Hence $\partial_{τ} u_{n} = - \frac{π i}{4} \partial_{z}^{2} u_{n}$ for every term, and summing over $n$ gives $\partial_{τ} θ_{3} = - \frac{π i}{4} \partial_{z}^{2} θ_{3}$ . The clean normalisation $\partial_{τ} θ = \frac{1}{4 π i} \partial_{z}^{2} θ$ appears in the rescaled convention with frequency $2 π n$ (term $e^{2 π in z}$ ); the two are related by $z \mapsto π z$ . $□$

Exercise 6 (hard, symbolic).

Assuming the Jacobi triple product $θ_{3} (z ∣ τ) = \prod_{n \geq 1} (1 - q^{2 n}) (1 + q^{2 n - 1} e^{2 i z}) (1 + q^{2 n - 1} e^{- 2 i z})$ , derive the product form of $θ_{1}$ .

Hint

$θ_{1} (z) = - i e^{i z} q^{1/4} θ_{3} (z + \frac{π}{2} (τ + 1) ∣ τ)$ ; track the quarter-shift through the product.

Answer

Shifting the argument of $θ_{3}$ by $\frac{π}{2}$ replaces $e^{2 i z}$ by $- e^{2 i z}$ in the product (since $e^{2 i (z + π /2)} = e^{2 i z} e^{π i} = - e^{2 i z}$ ), turning the factors $(1 \pm q^{2 n - 1} e^{\pm 2 i z})$ into $(1 \mp q^{2 n - 1} e^{\pm 2 i z})$ . Combining with the additional $\frac{π τ}{2}$ shift (which contributes a prefactor $e^{i z} q^{1/4}$ via quasi-periodicity) and the odd normalisation $- i$ , the surviving form is

θ_{1} (z ∣ τ) = 2 q^{1/4} sin z n \geq 1 \prod (1 - q^{2 n}) (1 - q^{2 n} e^{2 i z}) (1 - q^{2 n} e^{- 2 i z}) .

The factor $2 sin z = - i (e^{i z} - e^{- i z})$ supplies the simple zero of $θ_{1}$ at $z = 0$ and at every lattice point, consistent with $θ_{1}$ being the odd theta with one zero per fundamental cell. $□$

Exercise 7 (hard, symbolic).

Use the modular law $θ_{3} (0 ∣ - 1/ τ) = - i τ θ_{3} (0 ∣ τ)$ and the permutation $θ_{2} \leftrightarrow θ_{4}$ to show that the combination $(θ_{2} θ_{3} θ_{4}) (0 ∣ τ)$ transforms with weight $3/2$ under $τ \mapsto - 1/ τ$ .

Hint

Each nullwert picks up a factor $- i τ$ ; multiply three of them and use that the product is symmetric under exchanging $θ_{2}$ and $θ_{4}$ .

Answer

Under $τ \mapsto - 1/ τ$ : $θ_{3} (0) \mapsto - i τ θ_{3} (0)$ , while $θ_{2} (0 ∣ - 1/ τ) = - i τ θ_{4} (0 ∣ τ)$ and $θ_{4} (0 ∣ - 1/ τ) = - i τ θ_{2} (0 ∣ τ)$ (the permutation). Multiplying, $(θ_{2} θ_{3} θ_{4}) (0 ∣ - 1/ τ) = (- i τ)^{3} (θ_{4} θ_{3} θ_{2}) (0 ∣ τ) = (- i τ)^{3/2} (θ_{2} θ_{3} θ_{4}) (0 ∣ τ)$ , since the product is unchanged by the swap. The weight is therefore $3/2$ , and raising to the eighth power gives $Δ = 2^{- 8} (θ_{2} θ_{3} θ_{4})^{8}$ weight $12$ — the modular discriminant. $□$

Advanced results Master

Jacobi triple product. The identity

θ_{3} (z ∣ τ) = n = 1 \prod \infty (1 - q^{2 n}) (1 + q^{2 n - 1} e^{2 i z}) (1 + q^{2 n - 1} e^{- 2 i z})

factors the theta series into an infinite product. The proof identifies both sides as the unique (up to scale) holomorphic function with the prescribed quasi-periods and a single zero per cell at $z = π /2 + π τ /2$ , then fixes the constant by comparing the $q^{0}$ coefficient. Specialising $z$ recovers Euler's pentagonal number theorem and Gauss's identities; the product form makes the zeros of each theta function manifest, which the series does not. ^{[Mumford 1983]}

Theta-nullwerte and the quartic identity. Writing $ϑ_{j} = θ_{j} (0 ∣ τ)$ , the product formula gives

ϑ_{2} = 2 q^{1/4} n \geq 1 \prod (1 - q^{2 n}) (1 + q^{2 n})^{2}, ϑ_{3} = n \geq 1 \prod (1 - q^{2 n}) (1 + q^{2 n - 1})^{2}, ϑ_{4} = n \geq 1 \prod (1 - q^{2 n}) (1 - q^{2 n - 1})^{2} .

Jacobi's aequatio identica satis abstrusa (his own phrase) is the quartic relation

ϑ_{3}^{4} = ϑ_{2}^{4} + ϑ_{4}^{4},

a polynomial identity among the three nullwerte that is equivalent to the statement that the modular lambda function $λ = ϑ_{2}^{4} / ϑ_{3}^{4}$ satisfies $λ + (1 - λ) = 1$ with $1 - λ = ϑ_{4}^{4} / ϑ_{3}^{4}$ . This single identity is the analytic core of the level-two modular structure.

Relation to the Weierstrass $℘$ -function. The logarithmic second derivative of $θ_{1}$ recovers the elliptic function machinery of 06.01.25. Working with the lattice of periods $π, π τ$ ,

℘ (u ∣ τ) = - \frac{d ^{2}}{d u ^{2}} lo g θ_{1} (u ∣ τ) - \frac{π ^{2}}{3} \frac{θ _{1}^{'''} ( 0 )}{θ _{1}^{'} ( 0 )},

where the constant is fixed so that the Laurent expansion of $℘$ at $u = 0$ has no constant term. Because $θ_{1}$ has exactly one simple zero per cell, $\partial_{u}^{2} lo g θ_{1}$ has exactly one double pole per cell, matching $℘$ . The theta functions thereby supply a holomorphic "potential" whose curvature is the meromorphic $℘$ .

Modular discriminant and the eta function. Combining the nullwerte product formulas yields $ϑ_{2} ϑ_{3} ϑ_{4} = 2 η (τ)^{3}$ , where $η (τ) = q^{1/12} \prod_{n \geq 1} (1 - q^{2 n})$ is the Dedekind eta function. Hence the modular discriminant is

Δ (τ) = (2 π)^{12} η (τ)^{24} = 2^{- 8} (2 π)^{12} (ϑ_{2} ϑ_{3} ϑ_{4})^{8},

a weight- $12$ cusp form on $SL_{2} (Z)$ . The $j$ -invariant of 06.01.26 is then $j = E_{4}^{3} /Δ$ up to normalisation, tying the theta nullwerte to the entire modular hierarchy.

Synthesis. The four Jacobi theta functions are the foundational reason that elliptic and modular objects share one analytic substrate, and the central insight is that a single Gaussian lattice sum is simultaneously a section of a line bundle on the torus, a solution of the heat equation, and a half-integral-weight modular form. This is exactly the structure that the modular transformation law packages: Poisson summation is dual to lattice rescaling, the triple product is dual to the prescribed zero divisor, and the quartic identity is dual to the level-two covering of the modular curve. Putting these together, the relation $Δ = 2^{- 8} (ϑ_{2} ϑ_{3} ϑ_{4})^{8}$ generalises from genus one to the Riemann theta on Jacobians 06.06.05, and the bridge from $θ_{1}$ to $℘$ generalises the Weierstrass factorisation 06.01.25 to the modular setting. The pattern recurs in the elliptic genus 03.06.23, whose modularity is the theta transformation law in disguise, and in the Witten genus, where the theta substrate of loop-space index theory is the same Gaussian self-duality seen here.

Full proof set Master

Proposition (Quasi-periodicity multipliers of $θ_{1}$ ). The odd theta function satisfies $θ_{1} (z + π ∣ τ) = - θ_{1} (z ∣ τ)$ and $θ_{1} (z + π τ ∣ τ) = - q^{- 1} e^{- 2 i z} θ_{1} (z ∣ τ)$ .

Proof. From $θ_{1} (z) = - i \sum_{n} (- 1)^{n} q^{(n + 1/2)^{2}} e^{(2 n + 1) i z}$ , the shift $z \mapsto z + π$ multiplies the term of index $n$ by $e^{(2 n + 1) π i} = e^{π i} = - 1$ for every $n$ (since $2 n$ is even), giving $θ_{1} (z + π) = - θ_{1} (z)$ . For the second shift, $z \mapsto z + π τ$ multiplies the index- $n$ term by $e^{(2 n + 1) π i τ} = q^{2 n + 1}$ . The summand becomes $- i (- 1)^{n} q^{(n + 1/2)^{2} + 2 n + 1} e^{(2 n + 1) i z}$ . Now $(n + 1/2)^{2} + 2 n + 1 = n^{2} + n + 1/4 + 2 n + 1 = (n + 3/2)^{2} - 1 = (n^{'} + 1/2)^{2} - 1$ with $n^{'} = n + 1$ . Re-indexing $n \mapsto n^{'} = n + 1$ (which flips $(- 1)^{n}$ to $- (- 1)^{n^{'}}$ ) and factoring $q^{- 1}$ and the residual $e^{- 2 i z}$ from the shift of frequency yields $θ_{1} (z + π τ) = - q^{- 1} e^{- 2 i z} θ_{1} (z)$ . $□$

Proposition (Theta-multiplier consistency). The transformation $θ_{1} (z ∣ - 1/ τ) = - i (- i τ)^{1/2} e^{i τ z^{2} / π} θ_{1} (z τ ∣ τ)$ is consistent with the period and the $- 1/ τ$ rules generating a projective $SL_{2} (Z)$ action.

Proof. Apply the Poisson-summation argument of the Key Theorem to the full $z$ -dependent series rather than the nullwert. Completing the square in $π i τ (n + 1/2)^{2} + (2 n + 1) i z$ produces a Gaussian whose Fourier dual reproduces the series for $θ_{1}$ with $τ$ replaced by $- 1/ τ$ , $z$ rescaled to $z τ$ , and an overall Gaussian prefactor $e^{i τ z^{2} / π}$ together with the half-integral multiplier $(- i τ)^{1/2}$ . The factor $- i$ is fixed by matching $θ_{1}^{'} (0)$ on both sides. Composing this with $τ \mapsto τ + 1$ (which sends $q \mapsto - q$ and multiplies $θ_{1}$ by $e^{π i /4}$ ) realises the two generators $S = \begin{psmallmatrix}0&-1\\1&0\end{psmallmatrix}$ and $T = \begin{psmallmatrix}1&1\\0&1\end{psmallmatrix}$ . The associated multipliers form an eighth-root-of-unity system, so the action is projective; passing to $θ_{1}^{8}$ or to $Δ = 2^{- 8} (ϑ_{2} ϑ_{3} ϑ_{4})^{8}$ clears the multiplier and gives an honest weight- $12$ modular form. $□$

Connections Master

Weierstrass $℘$ -function 06.01.25. The function $℘$ is the negative second logarithmic derivative of $θ_{1}$ (after rescaling the lattice to periods $π, π τ$ ): the single simple zero of $θ_{1}$ per cell becomes the single double pole of $℘$ per cell. Where 06.01.25 builds elliptic functions from lattice sums with subtracted corrections, the theta presentation builds the same field from a holomorphic potential, and the addition formula for $℘$ becomes the three-term Riemann theta relation among shifted $θ_{1}$ values.
Modular function and $j$ -invariant 06.01.26. The theta-nullwerte assemble into the lambda function $λ = ϑ_{2}^{4} / ϑ_{3}^{4}$ , a Hauptmodul for the level-two congruence subgroup, and the $j$ -invariant is the degree-six rational function of $λ$ given by $j = 256 (1 - λ + λ^{2})^{3} / (λ^{2} (1 - λ)^{2})$ . The quartic identity $ϑ_{3}^{4} = ϑ_{2}^{4} + ϑ_{4}^{4}$ is exactly the relation $λ + (1 - λ) = 1$ that makes this rational map well-defined on the modular curve.
Modular forms on $SL_{2} (Z)$ 21.04.01. The transformation law derived here exhibits each theta-nullwert as a modular form of half-integral weight $1/2$ for a congruence subgroup, and the product $Δ = 2^{- 8} (2 π)^{12} (ϑ_{2} ϑ_{3} ϑ_{4})^{8} = (2 π)^{12} η^{24}$ is the weight- $12$ cusp form generating the ideal of cusp forms. Thus the theory of 21.04.01 inherits its first explicit cusp form from the theta machinery.
Elliptic genus 03.06.23. The Ochanine and Witten genera are built from the power series $x /2 \cdot θ_{1}^{'} (0) / θ_{1} (x /2)$ (up to normalisation), so the modularity of these genera is precisely the theta transformation law proved here applied with a formal variable. The rigidity theorem 03.06.24 consumes $θ_{1} (z / τ, - 1/ τ) = - i (- i τ)^{1/2} e^{iπ z^{2} / τ} θ_{1} (z, τ)$ as its key analytic input, which this unit supplies.
Riemann theta on Jacobians 06.06.05. The genus- $g$ Riemann theta $Θ$ is a multivariable generalisation summed over $Z^{g}$ with a period matrix in place of the single $τ$ ; in genus one it reduces to $θ_{3}$ . The two objects are different — capital $Θ$ is a function of $g$ complex variables and a symmetric period matrix, while the Jacobi thetas here are functions of one variable and one $τ$ — but they share the Poisson-summation transformation principle, generalised to the full symplectic group $Sp (2 g, Z)$ .

Historical & philosophical context Master

Jacobi 1829 ^{[Jacobi 1829]}, in Fundamenta Nova Theoriae Functionum Ellipticarum (Konigsberg), introduced the theta functions as the fundamental building blocks of elliptic-function theory, deriving the triple product and the quartic identity he memorably labelled aequatio identica satis abstrusa — "a rather abstruse identity." Jacobi's insight was that the awkward double periodicity of elliptic functions could be traded for the cleaner quasi-periodicity of theta functions, which are entire rather than meromorphic. This trade — meromorphic functions as ratios of holomorphic sections — is the historical seed of the modern theory of line bundles on abelian varieties.

The modular transformation $θ_{3} (0 ∣ - 1/ τ) = - i τ θ_{3} (0 ∣ τ)$ was already implicit in work on the heat equation and theta-series reciprocity; its derivation by Poisson summation became the template for the functional equations of Riemann's zeta function (via the theta-zeta integral) and of general $L$ -functions. Whittaker and Watson ^{[Whittaker-Watson Ch. 21]} codified the four-function notation that remains standard, and Mumford's Tata Lectures on Theta ^{[Mumford 1983]} recast the whole subject in the language of algebraic geometry, presenting the Jacobi thetas as the genus-one case of theta functions on abelian varieties and connecting them to the representation theory of the Heisenberg group. Chandrasekharan ^{[Chandrasekharan 1985]} gives the bridge from the nullwerte to the modular forms of level one.

Bibliography Master

@book{Jacobi1829Theta,
  author = {Jacobi, Carl Gustav Jacob},
  title = {Fundamenta Nova Theoriae Functionum Ellipticarum},
  publisher = {Borntr\"ager, K\"onigsberg},
  year = {1829},
  note = {Theta series, Jacobi triple product, aequatio identica satis abstrusa}
}

@book{WhittakerWatson1927,
  author = {Whittaker, E. T. and Watson, G. N.},
  title = {A Course of Modern Analysis},
  publisher = {Cambridge University Press},
  year = {1927},
  edition = {4th},
  note = {Chapter 21: the theta functions, quasi-periodicity and transformation}
}

@book{Mumford1983Tata,
  author = {Mumford, David},
  title = {Tata Lectures on Theta I},
  publisher = {Birkh\"auser},
  year = {1983},
  series = {Progress in Mathematics},
  volume = {28},
  note = {Chapter I: Jacobi theta, triple product, modular transformation via Heisenberg group}
}

@book{SteinShakarchi2003,
  author = {Stein, Elias M. and Shakarchi, Rami},
  title = {Complex Analysis},
  publisher = {Princeton University Press},
  year = {2003},
  note = {Chapter 10: theta functions, Poisson summation, sums of squares}
}

@book{Chandrasekharan1985,
  author = {Chandrasekharan, K.},
  title = {Elliptic Functions},
  publisher = {Springer-Verlag},
  year = {1985},
  series = {Grundlehren der mathematischen Wissenschaften},
  volume = {281},
  note = {Chapter V: theta-nullwerte, lambda function, and modular forms}
}

Prerequisites

06.01.25
06.01.26
21.04.01

Tier anchors

beginner: Stein-Shakarchi *Complex Analysis* (Princeton 2003) Ch. 10 (theta functions and applications); Needham *Visual Complex Analysis* (Oxford 1997) informal periodicity
intermediate: Whittaker-Watson *A Course of Modern Analysis* 4th ed. (Cambridge 1927) Ch. 21; Stein-Shakarchi *Complex Analysis* Vol. II Ch. 10; Ahlfors *Complex Analysis* (McGraw-Hill 1979) Ch. 7
master: Mumford *Tata Lectures on Theta I* (Birkhauser 1983) Ch. I; Jacobi 1829 *Fundamenta Nova Theoriae Functionum Ellipticarum* (Konigsberg); Whittaker-Watson Ch. 21; Chandrasekharan *Elliptic Functions* (Springer 1985) Ch. V

References

Jacobi 1829 — Fundamenta Nova Theoriae Functionum Ellipticarum · Konigsberg; Werke Vol. 1, theta series and product formula
Whittaker-Watson — A Course of Modern Analysis · Cambridge University Press 1927, 4th ed., Ch. 21 (theta functions)
Mumford — Tata Lectures on Theta I · Birkhauser 1983, Ch. I (Jacobi theta, transformation, triple product)
Stein-Shakarchi — Complex Analysis · Princeton University Press 2003, Ch. 10 (theta functions)
Chandrasekharan — Elliptic Functions · Springer 1985, Ch. V (Jacobi theta-nullwerte and modular forms)

Estimated time

beginner: 18m
intermediate: 45m
master: 85m