3
$\begingroup$

Denote by $\zeta$ the Riemann zeta function.

It is known that

$$\int_{0}^{T} \zeta(1/2 + it) \mathrm{d}t = T + O(T^{1/2}).$$

But is a similar result for $\int_{0}^{T} \zeta(\sigma + it) \mathrm{d}t$, where $0<\sigma<1$ also known ?

$\endgroup$
3
  • $\begingroup$ Should the integrand be the absolute value of zeta rather than zeta itself? $\endgroup$
    – Yemon Choi
    Dec 1, 2018 at 21:03
  • $\begingroup$ @Yemon Choi, the integrand is indeed zeta itself. $\endgroup$
    – sigma
    Dec 1, 2018 at 21:07
  • $\begingroup$ If you like my answer, please accept it officially (so that it turns green). Thanks in advance! $\endgroup$
    – GH from MO
    Dec 2, 2018 at 15:30

3 Answers 3

7
$\begingroup$

We have, for any $0<\sigma<1$, $$\int_0^T\zeta(\sigma+it)\,dt = T+O_\sigma(T^{1-\sigma}).$$ To see this, it suffices to show that $$\int_T^{2T}\zeta(\sigma+it)\,dt = T+O_\sigma(T^{1-\sigma}).$$ By Theorem 4.11 in Titchmarsh: The theory of the Riemann zeta-function, we have $$\zeta(\sigma+it)=\sum_{n\leq T}n^{-\sigma-it} + O_\sigma(T^{-\sigma}),\qquad T\leq t\leq 2T,$$ therefore by explicit integration $$\int_T^{2T}\zeta(\sigma+it)\,dt = T + \sum_{2\leq n\leq T}O(n^{-\sigma}) +O_\sigma(T^{1-\sigma}) = T+O_\sigma(T^{1-\sigma}).$$ We are done.

$\endgroup$
1
$\begingroup$

For $1/2<\sigma<1$, the same result should be known, and indeed should follow from the same proof: $\zeta(s)$ can be approximated by a suitably chosen truncation of its Dirichlet series (despite that series not converging inside the critical strip), and integrating term by term yields $T$ from the $1^{-s}$ term and $o(T)$ from all the $n^{-s}$ terms with $n\ge2$. (I don't know whether this method yields the strong error term cited, but the idea is worth sharing anyway: the asymptotic comes from the constant term in the Diriclet series.)

For $0<\sigma<1/2$, the answer will probably be different: for example, on the Lindelöf hypothesis, $|\zeta(\sigma+it)| \asymp t^{1/2-\sigma}$. So the integral will presumably have size around $T^{3/2-\sigma}$. I'm sure this is known (but not by me at this moment...).

$\endgroup$
1
  • 1
    $\begingroup$ If we have some aproximation $\zeta(\sigma + it) = \sum_n V(n) n^{-s}$ for smooth weight function $V$, the non-constant terms contribute $\sum_n V(n) i (n^{iT} - 1) /( n^\sigma \log n)$. If I remember correctly, $V$ should have height about one and width about $T$, which would make the nonconstant terms contribute at most $O ( T^{1 - \sigma})$. The main term is still $T$. I don't think Lindelöf is relevant. $\endgroup$
    – Will Sawin
    Dec 2, 2018 at 1:08
-2
$\begingroup$

What you say is incorrect. It is known that $$ \int_{0}^{T} \zeta(\tfrac 12 + it) dt = o(T). $$ Can you please correct the question, and ask again?

$\endgroup$
4
  • $\begingroup$ Im so curious now. Can you give a reference for this claim? $\endgroup$
    – BigM
    Dec 1, 2018 at 19:43
  • 2
    $\begingroup$ Well, i hadn't put a refernce because i assumed the result is fairly famous. Indeed, what i wrote in my question is known, see Montgomery-Vaughan, p. 458 $\endgroup$
    – sigma
    Dec 1, 2018 at 20:01
  • 1
    $\begingroup$ @wqeqwe, it is not true that $\int_{0}^{T} \zeta(1/2 + it) \mathrm{d}t =o(T)$. See Montgomery-Vaughan's ''Multiplicative Number Theory: The Classical Theory'', p.458. $\endgroup$
    – sigma
    Dec 1, 2018 at 21:11
  • $\begingroup$ wqeqwe: Lemma 14.10 in the book @sigma refers to is not consistent with your claim. (The main step in the proof establishes that $\int_1^T \zeta(1/2+it)\,dt = \int_1^T \zeta(2+it)\,dt + O(T^{1/2})$ and we can estimate the second integral easily using the Dirichlet series) $\endgroup$
    – Yemon Choi
    Dec 1, 2018 at 23:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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