ON THETA FUNCTIONS

BERNDT E. SCHWERDTFEGER

Abstract. Definition and properties of Jacobi’s ϑ–function.

Preface

In this paper I present the foundation of Jacobi’s ϑ–functions, based on his Notices sur
les fonctions elliptiques [3, vol. I, 7.] and his lecture Theorie der elliptischen Functionen
[3, vol. I, 19.]. I derive all his ϑ–relations, in particular his merkwürdige Relation of
theta–constants
ϑ400 (0, τ ) = ϑ401 (0, τ ) + ϑ410 (0, τ )

Berlin, 7 October 2007

c 2007 Berndt E. Schwerdtfeger, restored v1.0

1. Jacobi’s theta series

The source of Jacobi’s ϑ–functions is in his Notices sur les fonctions elliptiques
[3, vol. I, 7.] (Crelle, 1828), as well as his Fundamenta nova theoriae functionum
ellipticarum ([3, vol. I, 4.], 1829).
Later Jacobi reversed the development and started with the theta series to derive
the theory of elliptic functions in his lecture Theorie der elliptischen Functionen,
aus den Eigenschaften der Thetareihen abgeleitet prepared by Borchardt in 1838
on behalf of Jacobi [3, vol. I, 19.]. We will take a rapid walk through the first
part of Jacobi’s lecture. For his notations and comparison with later authors see
the section 5.
Let D = {x ∈ C | |x| < 1} be the open unit disk in C. The complex line C is the
universal covering of C× = C − {0} via the exponential map
e
0 → Z −→ C−→C× → 1
where t = e(x) = exp(2πix) ∈ C× for x ∈ C. In the following diagram

H /C

e e
  
D×  / C×

H = {τ ∈ C | Im τ > 0} = e−1 (D× ), the upper half plane, is the universal covering
of the puntured disk D× = D−{0}. Variables are denoted x ∈ C and t = e(x) ∈ C× ,
resp. τ ∈ H and (sic! ) q = e(τ /2) ∈ D× .

2000 Mathematics Subject Classification. Primary 14H42; Secondary 14K25.

Key words and phrases. theta function.
1
2 BERNDT E. SCHWERDTFEGER

2. Definitions and properties

Throughout this note I will use the following definition for ϑ

Definition 2.1. Let ϑ : C × H −→ C be given by:
X X 2
ϑ(x, τ ) = e(n2 τ /2 + nx) = q n · tn
n∈Z n∈Z

Remark. In Jacobi’s notation ϑ3 (πx, q) = ϑ(x, τ ), q = e(τ /2), see section 5.

Proposition 2.1. The series for ϑ converges absolutely and uniformly on compact
subsets, defining an analytic function
ϑ : C × H −→ C
(x, τ ) 7−→ ϑ(x, τ )

Proof. Let K ⊂ C × H be compact, then there is m, M ∈ R, m, M > 0 such that for

all (x, τ ) ∈ K, Im τ ≥ m and −M ≤ Im x ≤ M . Now let q0 = e−πm and t0 = e2πM ,
then |q| = |e(τ /2)| = exp(−π Im τ ) ≤ q0 and t−1
0 ≤ |t| ≤ t0 . We then have
( 2
τ 2 n2 n q0n · tn0 n≥0
|e( n + nx)| = |q| · |t| ≤ 2
2 q0n · t−n
0 n<0
n0
Take an integer n0 > 2M n
m , then q1 = q0 t0 < 1 and for n ≥ n0 we have q0 t0 ≤ q1 ,
2
n2 −n
hence q0 ·t0 = (q0 t0 ) ≤ q1 . Similarly for n ≤ −n0 , since −n ≥ n0 , q0 ·t0 ≤ q1−n .
n n n n n

Hence, the series is majorized by a geometric series on the compact set K. 

This proof of convergence can easily be adopted to the situation where we sum over
a shifted set a + Z, a ∈ C, instead of Z in the sum defining ϑ, leading to the
Definition 2.2. For a, b ∈ C the shifted ϑ is defined by
  X τ
a X τ
e (n + a)2 + (n + a)(x + b) = e n2 + n(x + b)



ϑ (x, τ ) =
b 2 2
n∈Z n∈a+Z

     
0 a a
So, in particular, ϑ(x, τ ) = ϑ (x, τ ) and ϑ (x, τ ) = ϑ (x + b, τ ).
0 b 0
Direct calculation yields the equation
 
a τ
(x, τ ) = e a2 + a(x + b) · ϑ(x + aτ + b, τ ),


(1) ϑ
b 2
which implies
   
a+m a
(2) ϑ (x, τ ) = e(an) · ϑ (x, τ ) for m, n ∈ Z,
b+n b
and in particular
(3) ϑ(x + 1, τ ) = ϑ(x, τ ),
τ
(4) ϑ(x + τ, τ ) = e(− − x) · ϑ(x, τ ) = q −1 t−1 ϑ(x, τ ).
2
We note that ϑ has a zero at x0 = (1 + τ )/2, as
X τ 1+τ X
ϑ(x0 , τ ) = e( n2 + n )= (−1)n q n(n+1) = 0.
2 2
n∈Z n∈Z
 ON THETA FUNCTIONS 3

τ τ +1It is in fact the only zero in a fundamental domain Dτ

of the lattice Λτ = Zτ + Z ⊂ C. This follows from the
Dτ x0
residue theorem: the number of zeros with multiplicity
0 1 is
ϑ0
Z 0
X X 1 ϑ (x, τ )
ordx ϑ(x, τ ) = Res( , x) = dx
ϑ 2πi γ ϑ(x, τ )
x∈Dτ x∈Dτ

where γ = ∂Dτ is the boundary of Dτ . We evaluate the integral over the path γ:
Z 0 Z 1 0 Z τ 0
ϑ (x, τ ) ϑ (x, τ ) ϑ0 (x + τ, τ )  ϑ (x + 1, τ ) ϑ0 (x, τ ) 
dx = − dx + − dx
γ ϑ(x, τ ) 0 ϑ(x, τ ) ϑ(x + τ, τ ) 0 ϑ(x + 1, τ ) ϑ(x, τ )
and the last integral is = 0 because of the periodicity (3). The logarithmic derivative
of (4) gives the relation
ϑ0 (x + τ, τ ) ϑ0 (x, τ )
= −2πi +
ϑ(x + τ, τ ) ϑ(x, τ )
P
and hence ordx ϑ(x, τ ) = 1, so there is exactly one simple zero of ϑ(x, τ ) at
x0 ∈ Dτ . 
For a, b ∈ 12 Z the following special notation is used.
 
a/2
Definition 2.3. For a, b ∈ {0, 1} define ϑab = ϑ .
b/2

By (1) this amounts to

X τ
e n2 + nx = ϑ(x, τ )


ϑ00 (x, τ ) =
2
n∈Z
X τ 1
(−1)n e n2 + nx = ϑ(x + , τ )


ϑ01 (x, τ ) =
2 2
n∈Z
X τ 1 1
τ x τ
ϑ10 (x, τ ) = e (n + )2 + (n + )x = e( )e( )ϑ(x + , τ )
2 2 2 8 2 2
n∈Z
X τ 1 1
τ x 1+τ
ϑ11 (x, τ ) = i(−1)n e (n + )2 + (n + )x = ie( )e( )ϑ(x + , τ)
2 2 2 8 2 2
n∈Z
1
We list a table of half period values1 ϑab (x + λ) for a λ ∈ 2 Λτ and where ε =
ε(x, τ ) = e(− x2 − τ8 ) is an exponential factor:
1 τ 1+τ
ϑ00 (x + ) = ϑ01 (x) ϑ00 (x + ) = εϑ10 (x) ϑ00 (x + ) = −iεϑ11 (x)
2 2 2
1 τ 1+τ
ϑ01 (x + ) = ϑ00 (x) ϑ01 (x + ) = −iεϑ11 (x) ϑ01 (x + ) = εϑ10 (x)
2 2 2
1 τ 1+τ
ϑ10 (x + ) = ϑ11 (x) ϑ10 (x + ) = εϑ00 (x) ϑ10 (x + ) = −iεϑ01 (x)
2 2 2
1 τ 1+τ
ϑ11 (x + ) = −ϑ10 (x) ϑ11 (x + ) = −iεϑ01 (x) ϑ11 (x + ) = −εϑ00 (x)
2 2 2
these follow from the definitions and (3) and (4) (cf. [3, vol. I, 19., (2.), p. 502]).
For completeness we also list the equations corresponding to (3) and (4):
ϑ01 (x + 1) = ϑ01 (x) ϑ01 (x + τ ) = −q −1 t−1 ϑ01 (x)
ϑ10 (x + 1) = −ϑ10 (x) ϑ10 (x + τ ) = q −1 t−1 ϑ10 (x)
ϑ11 (x + 1) = −ϑ11 (x) ϑ11 (x + τ ) = −q −1 t−1 ϑ11 (x)
1dropping τ from the notation
4 BERNDT E. SCHWERDTFEGER

3. Deriving a remarkable relation

In his 1838 lecture Jacobi proceeds to derive several formulas between sums of
products of four ϑ–series. To lighten the notation, let us agree that ϑ(x) =
ϑ(x1 )ϑ(x2 )ϑ(x3 )ϑ(x4 ) for vectors x = (x1 , x2 , x3 , x4 ) ∈ C4 and similarly for the
other thetas.
Jacobi considers the linear reflection at the hyperplane x1 − x2 − x3 − x4 = 0
C4 −→ C4
x 7−→ x0 = x · A
given by the matrix  
1 1 1 1
1 1 1 −1 −1 
A=  
2  1 −1 1 −1 
1 −1 −1 1
t 2
which satisfies A = A and A = 1. His Fundamentalsatz [3, vol. I, 19.2.,(11.)] is
the relation (5) in the next theorem:
Theorem 3.1.
(5) ϑ00 (x) + ϑ10 (x) = ϑ00 (x0 ) + ϑ10 (x0 )
(6) ϑ00 (x) − ϑ10 (x) = ϑ01 (x0 ) + ϑ11 (x0 )
(7) ϑ01 (x) + ϑ11 (x) = ϑ00 (x0 ) − ϑ10 (x0 )
(8) ϑ01 (x) − ϑ11 (x) = ϑ01 (x0 ) − ϑ11 (x0 )

Proof. Jacobi’s reasoning rests on the observation that for x, y ∈ C4 the bilinear
form x · t y = x1 y1 + x2 y2 + x3 y3 + x4 y4 is invariant under the involution x 7→ x0 :
x0 · t y 0 = x · A · t (y · A) = x · A · t A · t y = x · t y

With our convention for x ∈ C4 we note that by definition

X τ X τ
ϑ00 (x) = e( n · t n + n · t x) ϑ10 (x) = e( n · t n + n · t x)
4
2 1
2
n∈Z 4 n∈( 2 +Z)

Jacobi writes down the formulas relating n = (a, b, c, d) and n0 = (a0 , b0 , c0 , d0 )

1 1
a0 = (a + b + c + d) b0 = (a + b − c − d)
2 2
1 1
c0 = (a − b + c − d) d0 = (a − b − c + d)
2 2
4 1 4
and for n ∈ Z ∪ ( 2 + Z) he emphasizes that the numbers
a0 + b0 = a + b a0 + c0 = a + c a0 + d0 = a + d
are integers, which implies n0 ∈ Z4 ∪ ( 21 + Z)4 , hence the involution induces a
∼
bijection Z4 ∪ ( 12 + Z)4 −→ Z4 ∪ ( 21 + Z)4 on the index set. Together with the
invariances n · t n = n0 · t n0 , n · t x = n0 · t x0 , the relation (5) becomes obvious.
Applying (5) to (x1 + 1, x2 , x3 , x4 )0 = (x01 + 12 , x02 + 12 , x03 + 21 , x04 + 21 ) and making
use of the table of half period values we get (6). (7) is identical with (6), by
interchanging x with x0 , as x00 = x. (8) is obtained from (6) by applying the reverse
operation (x1 + 12 , x2 + 12 , x3 + 12 , x4 + 21 )0 = (x01 + 1, x02 , x03 , x04 ). 

The system of equations (5)–(8) can be combined into one vector equation

0
ϑ00 (x), ϑ01 (x), ϑ10 (x), ϑ11 (x) = ϑ00 (x0 ), ϑ01 (x0 ), ϑ10 (x0 ), ϑ11 (x0 )


(9)
 ON THETA FUNCTIONS 5

Remark. Of course, (9) and the system of equations (5)–(8) are equivalent. The
vector components of (9) correspond to the equations (R2)–(R5) of Mumford in
[5, I, §5], whereas (5)–(8) are identical to Jacobi’s table (A.): (1.)–(4.) in [3, vol.
I, 19., p. 507].

In the sequel Jacobi substitutes various different vectors into (9). I skip some of
them and only list the outcome for the vector (x, x, y, y)0 = (x + y, x − y, 0, 0)
ϑ00 (x + y)ϑ00 (x − y)ϑ200 (0) = ϑ200 (x)ϑ200 (y) + ϑ211 (x)ϑ211 (y) =
= ϑ201 (x)ϑ201 (y) + ϑ210 (x)ϑ210 (y)
ϑ01 (x + y)ϑ01 (x − y)ϑ201 (0) = ϑ200 (x)ϑ200 (y) − ϑ210 (x)ϑ210 (y) =
= ϑ201 (x)ϑ201 (y) − ϑ211 (x)ϑ211 (y)
ϑ10 (x + y)ϑ10 (x − y)ϑ210 (0) = ϑ200 (x)ϑ200 (y) − ϑ201 (x)ϑ201 (y) =
= ϑ210 (x)ϑ210 (y) − ϑ211 (x)ϑ211 (y)
For y = x in particular
ϑ00 (2x)ϑ300 (0) = ϑ400 (x) + ϑ411 (x) = ϑ401 (x) + ϑ410 (x)
whereas for y = 0 it gives ϑ200 (x)ϑ200 (0) = ϑ201 (x)ϑ201 (0)+ϑ210 (x)ϑ210 (0). Substituting
x 7→ x + 21 + τ2 yields −ε2 ϑ211 (x)ϑ200 (0) = ε2 ϑ210 (x)ϑ201 (0) − ε2 ϑ201 (x)ϑ210 (0), hence
ϑ200 (x)ϑ200 (0) = ϑ201 (x)ϑ201 (0) + ϑ210 (x)ϑ210 (0)
ϑ211 (x)ϑ200 (0) = ϑ201 (x)ϑ210 (0) − ϑ210 (x)ϑ201 (0)
Finally for x = 0 we obtain the remarkable relation (in Jacobi [3, vol. I, 19., (E.)
p. 511] die merkwürdige Relation) for the Theta–Nullwerte
(10) ϑ400 (0, τ ) = ϑ401 (0, τ ) + ϑ410 (0, τ )
i.e.
(1+2q+2q 4 +2q 9 +. . . )4 = (1−2q+2q 4 −2q 9 +. . . )4 +16q(1+q 1·2 +q 2·3 +q 3·4 +. . . )4

4. Variation with the module τ

In the previous section we have kept the module τ fixed (and sometimes dropped
it from the notation). We are now proving the behaviour of ϑ with respect to
variation of τ .
We start by looking at ϑ(x, τ + 1).
X τ +1 X τ 1
ϑ(x, τ + 1) = e( n2 + nx) = e( n2 + n2 + nx) =
2 2 2
X
n τ 2
= (−1) e( n + nx) = ϑ01 (x, τ ) = ϑ(x + 1/2, τ )
2
2
where we used (−1)n = (−1)n . In particular, ϑ(x, τ + 2) = ϑ(x, τ ).
Next we are going to look at ϑ(x, −1/τ ). Jacobi states in [3, vol. I, 7., p. 264] the
formula r    
K Kxx Kx 0
H(ix, q) = i exp H , q
K0 πK 0 K0
0
where q = exp( −πK 0 −πK 0
K ) and q = exp( K 0 ). With τ = iK /K (such that q = e(τ /2))
this can be rewritten (see section 5) in our notation as
p
ϑ11 (x, τ ) = i i/τ exp(−πix2 /τ )ϑ11 (x/τ, −1/τ )
This can be transformed into the following set of equivalent equations:
6 BERNDT E. SCHWERDTFEGER

Theorem 4.1.
p
ϑ(x, −1/τ ) = τ /i e(x2 τ /2)ϑ(xτ, τ )
p
ϑ01 (x, −1/τ ) = τ /i e(x2 τ /2)ϑ10 (xτ, τ )
p
ϑ10 (x, −1/τ ) = τ /i e(x2 τ /2)ϑ01 (xτ, τ )
p
ϑ11 (x, −1/τ ) = −i τ /i e(x2 τ /2)ϑ11 (xτ, τ )

Proof. The equivalence follows from the table of half period values. We are going
to prove the first one.
Recall some Fourier transforms: exp(−πx2 ) is its own transform, hence, for t > 0,
g(x) = exp(−πtx2 ) has the transform gb(x) = √1t exp(− πt x2 ) and for h(x) = g(x+a)
R R
h(x) = g(t + a)e(tx)dt = g(t)e((t − a)x)dt = gb(x)e(−ax).
we get b
P P
The Poisson formula h(n) = b h(n) now yields the equation
X 1 X π

exp − πt(n + a)2 = √ exp − n2 e(−an)


(11)
n
t n t
 
a
(0, τ ) = e( τ2 a2 )ϑ(aτ, τ ) = e τ2 (n + a)2 , which is the left hand
P

Now by (1) ϑ
0
side of our Poisson equation (11) for τ = ti, whereas the sum on the right hand
2
e − n2τ − an = ϑ(−a, −1/τ ) = ϑ(a, −1/τ ) and equation (11) reads
P

side is
τ p
e( a2 )ϑ(aτ, τ ) = i/τ ϑ(a, −1/τ )
2
which, by analytic continuation, holds for all τ ∈ H. 

5. Notation for theta functions by different authors

The notation in the literature varies. Jacobi himself used different notations for
his ϑ–functions at various times. In his Notices sur les fonctions elliptiques in
1828 in Crelle’s Journal [3, vol. I, 7.], Jacobi introduced the notation H (Eta)
and Θ (Theta) (loc.cit. p. 256) for the numerator resp. denominator of his sinus
amplitudinis
2Kx 1 H(x)
sin am =√
π k Θ(x)
√ p p p
H(x) = 2 q sin x − 2 4 q 9 sin 3x + 2 4 q 25 sin 5x − 2 4 q 49 sin 7x + . . .
4

Θ(x) = 1 − 2q cos 2x + 2q 4 cos 4x − 2q 9 cos 6x + 2q 16 cos 8x − . . .

A year later in his book Fundamenta Nova Theoriae Functionum Ellipticarum [3,
vol. I, 4., p. 198, p. 224] he used H( 2Kx 2Kx
π ) = H(x) and Θ( π ) = Θ(x) instead
(see the remarks of Weierstraß on p. 542). In the 1838 lecture they are called
ϑ1 (x) = H(x) resp. ϑ(x) = Θ(x), where his definitions are as follows:
For q ∈ C× such that |q| < 1 Jacobi defines (loc.cit. p. 501):
X 2 X 1 2
ϑ(x, q) = (−1)ν q ν e2νxi ϑ1 (x, q) = − i2ν+1 q 4 (2ν+1) e(2ν+1)xi
X 1 2 X 2
ϑ2 (x, q) = q 4 (2ν+1) e(2ν+1)xi ϑ3 (x, q) = q ν e2νxi

Later authors introduced other fashions, like θ versus ϑ, and ever changing sub-
scripts. Here is a translation table of notations by selected mathematicians.
 ON THETA FUNCTIONS 7

Jacobi Notices, 1828 Θ(πx) H(πx)

Jacobi Fundamenta nova, 1829 Θ(2Kx) H(2Kx)
Jacobi Lecture notes, 1838 ϑ(πx, q) ϑ1 (πx, q) ϑ2 (πx, q) ϑ3 (πx, q)
Weierstraß ϑ0 (x|τ ) ϑ1 (x|τ ) ϑ2 (x|τ ) ϑ3 (x|τ )
Hermite θ0,1 (x, τ ) −iθ1,1 (x, τ ) θ1,0 (x, τ ) θ0,0 (x, τ )
C. Jordan θ2 (x, τ ) θ(x, τ ) θ1 (x, τ ) θ3 (x, τ )
H. Cartan ϑ0 (x, τ ) ϑ1 (x, τ )
Mumford ϑ01 (x, τ ) −ϑ11 (x, τ ) ϑ10 (x, τ ) ϑ00 (x, τ )
Weierstraß in Einführung der Thetafunctionen P [6, §34.] describes the relation to
Jacobi and Hermite, who defined θµ,ν (x) = m (−1)mν e( τ2 (m+ µ2 )2 +(m+ µ2 )x).
Cartan follows Weierstraß in [1, chap. V, ex. 3]. Jordan introduced Les
fonctions θ(x, τ ), θ1 (x, τ ), θ2 (x, τ ), θ3 (x, τ ) (V, no 426) in Fonctions elliptiques
[4, chap. VII] and relates their difference to Weierstraß’ notation in no 427.
Chandrasekharan uses this notation in [2, V, §8] of Jordan’s, as does Weil in
[7, chap. IV, §8] for θ(ζ, τ ).
I have chosen Mumford’s notation in [5].

References
[1] Henri Cartan, Elementare Theorie der Analytische Funktionen einer oder mehrerer kom-
plexen Veränderlichen, Hochschultaschenbücher, vol. 112, Bibliographisches Institut, Man-
nheim, Wien, Zürich, 1966.
[2] Komaravolu Chandrasekharan, Elliptic functions, Grundlehren der math. Wiss., vol. 281,
Springer, Berlin, Heidelberg, New York, Tokyo, 1985.
[3] Carl Gustav Jacob Jacobi, Gesammelte Werke, AMS Chelsea Publishing, Providence, 1969.
[4] Camille Jordan, Cours d’Analyse, 3rd ed., Vol. II, Jacques Gabay, Paris, 1991. réimpression
autorisée de la 3. édition publiée par Gauthiers–Villars en 1913.
[5] David Mumford, Tata Lectures on Theta, Modern Birkhäuser Classics, Springer, 2007.
[6] Karl Weierstraß, Formeln und Lehrsätze zum Gebrauche der elliptischen Functionen, Göttin-
gen, 1885. Nach Vorlesungen und Aufzeichnungen des Herrn K. Weierstrass bearbeitet und
herausgegeben von H. A. Schwarz.
[7] André Weil, Elliptic Functions according to Eisenstein and Kronecker, Classics in Mathemat-
ics, Springer, Berlin, Heidelberg, New York, 1976, 1999.