2
$\begingroup$

I want to show that

$$ \lim_{s \to 1} \left( \zeta(s,a) - \frac{1}{s-1} \right) = - \psi(a)$$

where $\zeta(s,a)$ is the Hurwitz zeta function and $\psi(a)$ is the digamma function.

The only approach I can think of is to use the integral representation

$$\zeta(s,a) = 2 \int_{0}^{\infty} \frac{\sin (s \arctan \frac{t}{a} )}{(a^{2}+t^{2})^{s/2} (e^{2 \pi t}-1)} \, dt + \frac{1}{2a^{s}} + \frac{a^{1-s}}{s-1} $$ which is valid for all complex values of $s$.

$$ \begin{align} \lim_{s \to 1} \left( \zeta(s,a) - \frac{1}{s-1} \right) &= 2 \int_{0}^{\infty} \frac{\sin (\arctan \frac{t}{a} )}{(a^{2}+t^{2})^{1/2} (e^{2 \pi t}-1)} \, dt + \frac{1}{2a} \\ &= 2\int_{0}^{\infty} \frac{t}{(a^{2}+t^{2})(e^{2 \pi t}-1)} \, dt + \frac{1}{2a} \\ &= - 2 \, \frac{d}{d a} \int_{0}^{\infty} \frac{\arctan (\frac{t}{a})}{e^{2 \pi t}-1} \, dt + \frac{1}{2a}.\end{align}$$

Binet's second log gamma formula states

$$ 2 \int_{0}^{\infty} \frac{\arctan \left( \frac{x}{z} \right)}{e^{2 \pi x} -1} \, dx = \ln \Gamma(z) - \left( z- \frac{1}{2} \right) \ln z + z - \frac{\ln (2 \pi)}{2}.$$

So

$$ \begin{align} \lim_{s \to 1} \left( \zeta(s,a) - \frac{1}{s-1} \right) &= - \frac{d}{da} \left( \ln \Gamma(a) - \left( a- \frac{1}{2} \right) \ln (a) + a - \frac{\ln (2 \pi)}{2} \right) + \frac{1}{2a} \\ &= -\psi(a) +\ln (a) +1 -\frac{1}{2a} -1 + \frac{1}{2a} \\ &= - \psi(a) \color{#C00000}{+ \ln (a)} . \end{align} $$

Why do I have $\ln (a)$ in my final answer?

$\endgroup$
2
$\begingroup$

Everything is correct except in the beginning where you take a limit of the integral representation. You conclude that $\lim_{s\rightarrow 1}\frac{a^{1-s}-1}{s-1}=1$, whereas by l'Hopital's rule, the correct answer is $-\ln(a)$ which should cancel the other log that you get later.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.