Stieltjes constants of L-functions in the extended Selberg class

Let f be an arithmetic function and let \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {S}}^\#$$\end{document}S# denote the extended Selberg class. We denote by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {L}}(s) = \sum _{n = 1}^{\infty }\frac{f(n)}{n^s}$$\end{document}L(s)=∑n=1∞f(n)ns the Dirichlet series attached to f. The Laurent–Stieltjes constants of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {L}}(s)$$\end{document}L(s), which belongs to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {S}}^\#$$\end{document}S#, are the coefficients of the Laurent expansion of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {L}}$$\end{document}L at its pole \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$s=1$$\end{document}s=1. In this paper, we give an upper bound of these constants, which is a generalization of many known results.


Introduction
Let q be any positive integer ≥ 1 and let χ be a Dirichlet character modulo q. Let γ n (χ ) denote the Laurent coefficients of the Dirichlet L-function L(s, χ) near s = 1. We recall that γ n (χ ) = In particular, γ 0 (1, 1) = 0.5772156649 · · · is the well-known Euler constant. The constants γ n (a, q) are often called the Stieltjes constants or generalized Euler constants. In the particular case when χ = χ 0 , where χ 0 is the principal character modulo 1, the Dirichlet L-function L(s, χ 0 ) reduces to the Riemann zeta function ζ(s), that is L(s, χ 0 ) = ζ(s). We write the corresponding Laurent coefficients simply γ n (χ 0 ) = γ n (1, 1) = γ n . Stieltjes in 1885 showed that which pioneered the study of Laurent coefficients of zeta functions and L-functions. This gives rise to the widely used name "Stieltjes constants" for these coefficients. The asymptotic behavior of γ n as n → ∞ has been widely studied by many authors (for instance: Briggs [5], Mitrovic [13], Israilov [10], Matsuoka [12], and more recently, Coffey [6,7], Knessl and Coffey [11], Adell [2], Adell and Lekuona [1], and Saad Eddin [16]). The studies mostly focused on the growth and sign changes of the sequence (γ n ), explicit upper estimates for |γ n |, and asymptotic expressions for γ n . Stieltjes constants for other zeta functions and L-functions have also been studied by many authors. We introduce some of their results in the following section.
In this paper, we are interested in investigating the Stieltjes constants of more general L-functions. We consider functions in a class larger than the Selberg class. We first introduce the Selberg class S.

Selberg class
Let f be an arithmetic function, and denote by L(s) = ∞ n=1 f (n) n s the Dirichlet series attached to f . We say the Dirichlet series L(s) belongs to the Selberg class S if it is absolutely convergent when Re(s) > 1 and satisfies the following properties: Condition S 1 Ramanujan hypothesis: For any ε > 0, we have f (n) ε n ε . Condition S 2 Analytic continuation: There exists k ∈ Z ≥0 such that (s − 1) k L(s) is entire of finite order.

Condition S 3 Functional equation: Define
where Q, λ i are positive real numbers, is the gamma function, μ j is a complex number satisfying Re(μ j ) ≥ 0. Then the function (s):=F(s)L(s) satisfies the functional equation where ω is a complex number with |ω| = 1.
Condition S 4 Euler product: For Re(s) > 1, the function L(s) can be written as a product over prime numbers p: with b(n) n θ , for some θ < 1 2 .
All functions L in S are expected to satisfy the Riemann hypothesis. If we remove any of the above conditions, however, this is no longer true. In this particular case, the hypothesis is usually called the Grand Riemann Hypothesis: all non-trivial zeros of L(s) ∈ S lie on Re(s) = 1/2. The extended Selberg class S # is defined to be the class of functions L(s) = ∞ n=1 f (n) n s satisfying the above conditions S2and S3, but not necessarily S1 and S4.
Notable examples of functions in S are the Riemann zeta function ζ(s), Dirichlet L-functions L(s, χ) associated with non-principal primitive characters χ , and the Dedekind zeta function ζ K (s) of a number field K . The sum of the parameters λ j in S3 gives the degree of the L-function L(s) in S # , and so in S, as follows: It is not known if d L ∈ Z >0 for all L ∈ S but the degree d L characterizes certain properties of the functions L ∈ S. Although the functional equation is not unique, the degree d L is well defined and captures the growth of L. We shall not discuss d L further since it is irrelevant to the aim of this paper. The readers may refer to [19,Chap. 6] for more details about the Selberg class.
From now on we keep our focus on L ∈ S # . That is, we would like to extend beyond the Selberg class by eliminating conditions S1 and S4. Consider the Laurent expansion of L(s) at its possible pole s = 1 written in the following form: We call the coefficients γ n (L) the generalized Laurent-Stieltjes constants or the Laurent-Stieltjes constants of the extended Selberg class. In this paper, we study these coefficients and give an upper bound of γ n (L) for L ∈ S # .
Our main theorem is stated as follows.
Theorem Let L ∈ S # \ {0} and let d L be the degree of L. Let Q be the positive real number appearing in condition S3 and let For a positive integer n with we have Finally we remark that the Laurent-Stieltjes constants of zeta and L-functions have many applications not only in analytic number theory, but also in algebraic number theory and even fields outside of number theory. They can be used to determine zerofree regions for L(s, χ) near the real axis in the critical strip 0 ≤ Re(s) ≤ 1, to compute the values of ζ(s) in the complex plane, to study the class number of a quadratic field, etc. Further, they appear in an explicit expression of Li's criterion which relates them to the Riemann hypothesis, see [4].

Some known results on the Laurent-Stieltjes constants of zeta and L-functions
The first explicit upper bound for |γ n | has been given by Briggs [5], which was later improved by Berndt [3] and Israilov [10]. In 1985, the theory made a huge progress via an asymptotic expansion shown by Matsuoka [12], for these constants. He gave an excellent upper bound of |γ n | for n ≥ 10 and proved that |γ n | ≤ 10 −4 e n log log n .
This result had been the best upper bound of |γ n | for more than 20 years. Thanks to the above result, Matsuoka showed that ζ(s) has no zeros in the region |s − 1| ≤ √ 2, with 0 ≤ Re(s) ≤ 1. Many have tried to improve on the Matsuoka bound, with few successful attempts. Matsuoka's work relied on a formula that is essentially a consequence of Cauchy's integral theorem and the functional equation. More recently, the second author [16] extended this formula to Dirichlet L-functions. She gave the following upper bound for |γ n (χ )| for primitive Dirichlet characters χ modulo q and for every 1 ≤ q ≤ π e (n+1)/2 /(2n + 2). We have In the case when χ = χ 0 and q = 1, this leads to a sizable improvement of Matsuoka's bound and of previous results. As an application of this upper bound, the second author showed in [17] that this result enables us to approximate L(s, χ) in the neighborhood of s = 1 by a short Taylor polynomial. For N = 4 log q and q ≥ 150, we have where |s − 1| ≤ e −1 . She also proved that the function ζ(s) has no zeros in the region |s − 1| ≤ 2.2093 with 0 ≤ Re(s) ≤ 1. This result is an improvement of Matsuoka's result.
Finally, let K be a number field and O K be its ring of integers. Define for Re(s) > 1 the Dedekind zeta function where a runs over non-zero ideals in O K , p runs over the prime ideals in O K and N a is the norm of a. It is known that ζ K (s) can be analytically continued to C \ {1}, and that at s = 1 it has a simple pole, with residue γ −1 (K ) given by the analytic class number formula: Here we denote by r 1 the number of real embeddings of K , r 2 the number of complex embeddings of K , h(K ) the class number of K , R(K ) the regulator of K , ω(K ) the number of roots of unity contained in K and D(K ) the discriminant of the extension K /Q. Consider the Laurent expansion The constants γ n (K ) are sometimes called the Stieltjes constants associated with the Dedekind zeta function. In [8] they are called higher Euler's constants of K . The second author [18] studied these constants and showed that, for n ≥ 1, we have To conclude this section, we remark that only the first constant γ K = γ 0 (K )/γ −1 (K ), called the Euler-Kronecker constant, which is closely related to values of the logarithmic derivative of L-functions, has been studied so far. For more details see for example [9,14,20]. This raises questions on the other Stieltjes constants associated with ζ K (s). The authors were motivated to give partial answers to these questions in a much more general context, that is, for all L-functions in the extended Selberg class.

Auxiliary lemmas
In order to prove our main result, we first show a proposition and two necessary lemmas. Recall the notation used when we defined S # in Sect. 1.

Lemma 1
Let L ∈ S # and let d L be the degree of L. Then we have as |t| → ∞. In particular, Proof For the standard case when we assume S1, see [19,Theorem 6.8]. Note here that the first-half (2) is obtained from the functional equation S3. Now without S1, we note that the absolute convergence of the Dirichlet series and partial summation imply for any ε > 0, Note that this is weaker than S1, but is sufficient for our purpose. Using (4) in combination with the functional equation S3, we can easily show the bounds (3) for the case σ ≥ 1 + ε and σ ≤ −ε. More precisely, Since the function L(s) is entire of finite order from condition S2, for any δ > 0, where the function G L is defined by Here Q, λ i are positive real numbers, μ j and ω are complex numbers with Re(μ j ) ≥ 0 and |ω| = 1.
Proof By Cauchy's formula, we can write where D is the positively oriented rectangular path passing through the vertices −a + 1+i T , −a +1−i T , A−i T and A+i T , where A and T are sufficiently large numbers. Let us now divide D into the line segments D 1 , D 2 , D 3 and D 4 joining −a + 1 + i T , −a + 1 − i T , A − i T , A + i T and −a + 1 + i T , as in Fig. 1. Then, we have γ n (L) = 1 2πi By Lemma 1, the integral over D 2 is bounded by Since a < n+1 d L + 1 2 , the last term vanishes as T → +∞. Therefore, the integral over D 2 tends to 0 as T → +∞. A similar argument shows that the integral over D 4 tends to 0 as T → +∞.
Next we consider the integral over D 3 . For n > 0, we find that Thus, for n > max 0, d L 2 − 1 , we have Here, by using the functional equation S3 for L(s) and the formula (s) (1 − s) sin(π s) = π , we have where the function G L (s) is as defined in (5). This completes the proof of Proposition 1.

Lemma 2
For L ∈ S # \ {0}, consider G L as defined in Proposition 1. Let λ m := min 1≤ j≤r λ j and μ M := max 1≤ j≤r |μ j |. For a > 1 + μ M λ m , we have We can easily show that On the other hand, using Stirling's formula we can show that, for x > 0, where the function ϕ(x, y) satisfies the inequality (cf. Binet's first formula) .
From these inequalities, we find that (note that Re(λ j (a − 1 + it) .

Hence we have
, which completes the proof.

Proof of Theorem
Now we are ready to prove our main theorem. We again put and let a be a real number satisfying 1 + μ M +1 λ m < a < 1 2 + n d L . By Proposition 1 and Lemma 2, we have We divide the region of integration into two as follows: (7) with A = a + 1 λ M . We estimate J 1 and J 2 in the following manner: and Substituting the above into (7), we obtain Therefore putting for 1 + μ M +1 λ m < a < 1 2 + n d L , we obtain |γ n (L)| ≤ C L (a)a −n 2 + 1 which completes the proof of our Theorem.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.