MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The coefficient of Selberg Class L-function satisfy:

$a_n <M_{\epsilon} n^{\epsilon}$ (for any $\epsilon >0$) and the $a_n$ are multiplicative.

So I would like to know if it can be shown that we also have the following partial sum bounded by a constant (maybe using also other properties of $a_n$ coefficients):

$\sum_{n=1}^{2N} a_n - a_2 a_n < M$

(As $a_n$ are multiplicative, we see that some terms disappear from the sum as $a_{2k}=a_2 a_k$ for k odd, but is there a chance to have this sum bounded by a constant?)

More generally what can you advise to read on the characterisation of coefficent of Selberg Class L-function?

share|cite|improve this question
Bertrand, if you like my answer (after studying Landau's oscillation theorem), then please accept it officially. Thanks! – GH from MO Sep 10 '13 at 12:45
up vote 3 down vote accepted

I think this fails for the Dirichlet coefficients of $L(s):=\zeta(s+i\gamma)\zeta(s-i\gamma)$ whenever $\gamma>0$ satisfies $2^{i\gamma}+2^{-i\gamma}\neq 1$. Indeed, for such $\gamma$, the inequality would imply that $S(x):=\sum_{n\leq x}a_n$ is bounded by a constant either from below or from above. However, $L(s)$ has simple poles at $1\pm i\gamma$ and it is holomorphic elsewhere, hence a quantitative form of Landau's oscillation theorem yields $S(x)=\Omega_{\pm}(x)$. See Theorem 11.8 in Bateman-Diamond: Analytic number theory - An introductory course.

share|cite|improve this answer
Thanks, it helps (i will need to read carrefully your reference) – Bertrand Sep 10 '13 at 12:29
In fact a characterisation like, they are constants A and M such that $|\sum_{n=1}^{2N} a_n - A \sum_{n=1}^{N} a_{2n}|<M$ seems more reasonable. – Bertrand Sep 10 '13 at 12:50
@Bertrand: I think your second proposed inequality fails already for $\zeta(s)^2=\sum a_nn^{-s}$. In this case the sum of $a_n$ over the odd (resp. even) integers up to $x$ is of the form $(1/4)x\log x+c_1 x$ (resp. $(3/4)x\log x+c_2 x$) plus lower order terms, where $c_1$ and $c_2$ are linearly independent over $\mathbb{Q}$ (I haven't checked this). So no $A$ is capable of neutralizing the $x\log x$ and $x$ terms simultaneously. – GH from MO Sep 10 '13 at 14:23
Yes your example is very interesting, thanks for your help, showing there is no such dependance between odd and even coefficients. – Bertrand Sep 11 '13 at 12:55

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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