[Math] Proving $\frac{\sin\pi z}{\pi z}=\prod_{n=1}^{\infty}\Big(1-\frac{z^{2}}{n^{2}}\Big)$

complex-analysisintegration

I apoligize if this has been answered already; the quick searches I've done have proven fruitless.

I'm given that

$\displaystyle\pi\cot(\pi z)=\frac{1}{z}+\sum_{1}^{\infty}\frac{2z}{z^{2}-n^{2}}$

and wish to derive

$\displaystyle\frac{\sin\pi z}{\pi z}=\prod_{n=1}^{\infty}\Big(1-\frac{z^{2}}{n^{2}}\Big)$

from this. I have taken the first equation, integrated both sides (using a u-substitution on the left-hand side with $u=\sin(\pi z)$ ) and come up with

$\displaystyle\ln(\sin(\pi z))=\ln(x)+\sum_{n=1}^\infty\ln(z^2-n^2)=\ln\Big(z\prod_{n=1}^\infty (z^2-n^2)\Big)$ which yields $\displaystyle\sin(\pi z)=z\prod_{n=1}^\infty(z^2-n^2)$, but the product here doesn't converge. Where exactly am I going wrong? Thanks in advance!

Best Answer

You need to be more careful about your antiderivatives: firstly, the cotangent identity shows that $$ \pi \cot{\pi z} - \frac{1}{z} $$ is finite (and hence integrable) in a neighbourhood of the origin (use the comparison test or the integral test or something). Hence, $$ \int_0^z \left(\pi \cot{\pi w} - \frac{1}{w}\right) \, dw \tag{1} $$ exists: an antiderivative is $$ \log{\left( \frac{\sin{\pi w}}{\pi w} \right)}, $$ so the integral (1) is $$ \left[ \log{\left( \frac{\sin{\pi w}}{z} \right)} \right]_0^z = \log{\left( \frac{\sin{\pi z}}{\pi z} \right)} - \log{1} = \log{\left( \frac{\sin{\pi z}}{\pi z} \right)}. $$

Now, for the right-hand side, we observe that it is uniformly convergent on $[0,z]$ for $z$ small enough, so we can integrate term-by-term. We have $$ \int_0^z \frac{2w}{w^2-n^2} \, dw, $$ and as you almost note, $\log{(n^2-w^2)}$ is an antiderivative (note the sign: $z<1$, so $z<n$). However, if we now evaluate the integral, we find $$ \left[ \log{(n^2-w^2)} \right]_0^z = \log{(n^2-z^2)}-\log{(n^2)} = \log{\left( 1-\frac{z^2}{n^2} \right)} $$ Summing these up, exponentiating and multiplying by $\pi z$, we get $$ \sin{\pi z} = \pi z \prod_{n=1}^{\infty} \left( 1-\frac{z^2}{n^2} \right) $$ on $[-1+\varepsilon, 1-\varepsilon]$, for any $\varepsilon>0$, because we can only get uniform convergence here.

The remaining issue is periodicity: you have to check that the right-hand side has period $2$. Extending the identity over the integers is straightforward by taking $\varepsilon \to 0$ and checking continuity there.

Related Question