I am actually aware of the argument showing $\zeta$ has a meromorphic extension to $\mathbb{C}$ with a single pole at $z = 1$. On a recent number theory exam, however, one of the questions asked to show $\zeta$ has a meromorphic extension to $\mathrm{Re} \; z > -1$ (and to show the same for the Hurwitz zeta function). Since this was a timed exam with multiple problems I doubt the professor was looking for the argument mentioned earlier. Instead I think there must be a comparatively elementary way of showing this, which leads to my question:

Is there a relatively simple way to prove that $\zeta$ has a meromorphic extension to $\mathrm{Re} \; z > -1$ with only a simple pole at $z = 1$?

If it helps, the first part of this problem on the exam asked us to establish the following: For a continuously differentiable function $f$ on $\{x \in \mathbb{R} : x > 0\}$ there is an equality

$$\sum_{k=1}^n f(k) = \int_1^n f(x) dx + \frac{1}{2}(f(1)+f(n)) + \int_1^n \left(x - [x] -\frac{1}{2}\right)f'(x) dx$$

Not sure if this last piece of information helps or is simply a red herring, but I figured it would be best to include it.


Yes, your instinct was right to include the first part of the problem! The left-hand side of the equality becomes $\zeta(s)$ when you set $f(k)=k^{-s}$ and let $n$ go to infinity; so let's do the same to the right-hand side.

But first: let $B_2(y) = \int_1^y (x-\lfloor x\rfloor-\frac12)\,dx$. Note that $B_2$ is a continuous periodic function with period $1$ (taking the value $0$ at every positive integer) and in particular is bounded. Note that by integration by parts, $$ \int_1^n \bigg(x - [x] -\frac{1}{2}\bigg)x^{-s-1} \,dx = (s+1) \int_1^n B_2(x) x^{-s-2}\,dx. $$

So now: when Re $s>1$, we have \begin{align*} \zeta(s) &= \lim_{n\to\infty} \sum_{k=1}^n k^{-s} \\ &= \lim_{n\to\infty} \bigg( \int_1^n x^{-s} \,dx + \frac{1}{2}(1^{-s}+n^{-s}) - s \int_1^n \bigg(x - [x] -\frac{1}{2}\bigg)x^{-s-1} \,dx \bigg) \\ &= \int_1^\infty x^{-s} \,dx + \frac{1}{2} - s \lim_{n\to\infty} (s+1) \int_1^n B_2(x) x^{-s-2}\,dx \\ &= \frac1{s-1} + \frac{1}{2} - s(s+1) \int_1^\infty B_2(x) x^{-s-2}\,dx, \tag{$*$} \end{align*} provided the integral converges. But in fact it converges absolutely, since $|B_2(x)|$ is bounded by some $C$: $$ \int_1^\infty |B_2(x) x^{-s-2}|\,dx \le C \int_1^\infty x^{-s-2}\,dx = \frac C{s+1}. $$

This gives the representation ($*$) for $\zeta(s)$, which we derived assuming that Re $s>1$. However, the right-hand side of ($*$) is convergent and well-defined for Re $s>-1$ (by the argument we just did). Therefore ($*$) gives an analytic continuation of $\zeta(s)$ to the region Re $s>-1$.

If this seems a little magical (including my choice to call that function $B_2$), you can read about Euler-Maclaurin summation to get some context.

  • $\begingroup$ Thanks! So yes it definitely appears my professor was aiming for this elementary proof. $\endgroup$ – James Miller Mar 10 '14 at 0:19

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.