1. Weierstrass ζ-function
Let ω1, ω2 be two complex numbers such that Im τ > 0 where τ = ω2/ω1. Let L = Zω1⊕ Zω2 be the lattice generated by {ω1, ω2}. The Weierstrass ζ function is defined by
ζ(z) = 1
z + X
ω∈L\{0}
1
z − ω+ 1 ω + z2
ω2
It follows from the definition of ζ and ℘(z) that ζ0(z) = −℘(z).
Since ℘(z) is an even function, ζ is an odd function, i.e. ζ(−z) = −ζ(z). If we write
℘(z) = 1 z2 +
∞
X
n=1
bnz2n in a neighborhood of 0, then
ζ(z) = 1 z−
∞
X
n=1
bn
2n + 1z2n+1
in a neighborhood of 0. By integrating the relation ℘(z + ω1) = ℘(ω), we obtain ζ(z + ω1) = ζ(z)+C, where C is a constant to be determined. Substituting z = −ω1/2 into this equation and using ζ(−z) = −ζ(z), we find C = 2ζ(ω1/2). Hence we obtain ζ(z+ω1) = ζ(z)+2ζ ω21 . In fact, we have
ζ(z + ωi) = ζ(z) + 2ζωi 2
, i = 1, 2, 3, where ω3= ω1+ ω2. We denote
ηi = ζωi 2
, i = 1, 2, 3.
Theorem 1.1. The Weierstrass ζ-function is an odd function of z such that ζ0(z) = −℘(z) and ζ(z + ωi) = ζ(z) + 2ηi for i = 1, 2, 3. It is holomorphic on C \ Λ and has pole at ω ∈ Λ of residue 1.
Theorem 1.2. (Legendre relation)
η1ω2− η2ω1 = πi.
Let us for simplicity denote Q
ω∈Λ\{0} by Q0
ω and P
ω∈Λ\{0} by P0
ω. Let us define E(z) = (1 − z)ez+z22 , z ∈ C.
Lemma 1.1. For |z| ≤ 1/2,
|E(z) − 1| ≤ 2|z|3. This lemma implies that the finite product
Y
ω 0E
z ω
converges absolutely and uniformly on |z| ≤ R for every R > 0. This allows us to define an entire function, called the Weierstrass σ-function by
σ(z) = z Y
ω
0Ez ω
= z Y
ω 0
1 − z ω
eωz+2ωz2.
1
2
After taking the principal branch of log, it follows from the definition that d
dzlog σ(z) = ζ(z).
The σ function is an odd function of z. Integrating ζ(z + ωi) = ζ(z) + 2ηi, we obtain σ(z + ωi) = ciσ(z)e2ηiz
where ci are constant to be determined. Substituting z = −ωi/2 into the equation and using σ being odd, we find ci = −e−2ηiωi. In other words, we obtain
(1.1) σ(z + ωi) = −σ(z)e2ηi(z+ωi2 )
Theorem 1.3. Let f (z) be an elliptic function of period {ω1, ω2}. Let a1, · · · , an be the zeros and b1, · · · , bn be the poles of f (z) repeated according to their multiplicity which lie in a fundamental period parallelogram P so that
n
X
i=1
ai ≡
n
X
j=1
bj mod L.
Then there exists a constant c ∈ C such that for all z ∈ C, (1.2) f (z) = c · σ(z − a1) · · · σ(z − an)
σ(z − b1) · · · σ(z − bn).
Proof. Let g(z) be the meromorphic function defined by the right hand side of equation (1.2). Equation (1.1) implies
g(z + ω1) = (−1)ne2η1(nz+n2ω1−a1−···−an)σ(z − a1) · · · σ(z − an) (−1)ne2η1(nz+n2ω1−b1−···−bn)σ(z − b1) · · · σ(z − bn)
= g(z).
Similarly, we have g(z + τ ) = g(z). This shows that g(z) is an elliptic function of periods {1, τ }. The quotient f (z)/g(z) is again elliptic without poles. By Lemma ??, f (z)/g(z) is a constant function. This completes the proof of our assertion.