The sum of the zeros and of the poles in a period parallelogram of an elliptic function f is a period of f

525 Views Asked by At

Let $f$ be an elliptic function with a fundamental set of periods $\omega_1, \omega_2$ (so their ratio is non-real).

Let $S$ be the sum of its poles and zeros inside a period parallelogram.

My attempt:

If $f=\prod_i (z-a_i)^{\alpha_i}$, then $\frac{f'}{f}=\sum_i \frac{\alpha_i}{z-a_i}$ and $\frac{1}{2i\pi}\int_c z\frac{f'}{f}(z)dz=\sum a_i\alpha_i=S$

$S$ is a period iff $S=n \omega_1 + m\omega_2$ for some integers $n,m$

Now, $g(z)=\frac{f'}{f}(z)$ is also elliptic so if I picture a parallelogram ABCD, then obviously $\int_{AB+CD} zg(z)dz=\int_{AB} zg(z)dz+\int_{CD} zg(z)dz$

Since ABCD is a period parallelogram, then $z_{CD}=z_{AB}+\omega_1$, and $g$ is elliptic so,

$\int_{AB+CD} zg(z)dz=\int_{AB} zg(z)dz-\int_{AB} (z+\omega_1)g(z)dz=-\omega_1\int_{AB}g(z)dz$

A similar reasoning yields a similar result for BC and $\omega_2$.

My problem is that I can't prove that $\frac{1}{2i\pi}\int_{AB}g(z)dz$ is an integer. The only things I know about $g$ is that it is elliptic, its integral over the whole parallelogram is $0$ and it counts the difference between the number of zeros and poles inside it. But this is not a closed contour so it is not so useful.