Abstract
In this paper we review the study of the distribution of the zeros of certain approximations for the Ramanujan \(\Xi \)-function given by Ki (Ramanujan J 17(1):123–143, 2008), and we provide new proofs of his results. Our approach is motivated by the ideas of Velásquez (J Anal Math 110:67–127, 2010) in the study of the zeros of certain sums of entire functions with some condition of stability related to the Hermite–Biehler theorem.
Similar content being viewed by others
1 Introduction
1.1 Background
Let \(\tau (n)\) be the Ramanujan’s tau-function, defined by
where \(q=e^{2\pi iz}\), and \(\mathrm{Im}\,{z}>0\). It is well known that \(\Delta (z)\) spans the space of cusp forms of dimension \(-12\) associated with the unimodular group. The associated Dirichlet series and Euler product for \(\Delta (z)\) are given by
where the series and the product are absolutely convergent for \(\mathrm{Re}\,{s}>13/2\).
Let us define the Ramanujan \(\Xi \)-function, denoted by \(\Xi _R(s)\), as follows:
where \(\Gamma (s)\) is the Gamma function. Another representation for \(\Xi _R(s)\) is given by
where
In [1], Hardy highlighted the importance of the location of the zeros of \(\Xi _R(s)\) in the strip \(|\mathrm{Im}\,(s)|\le \tfrac{1}{2}\). The Riemann hypothesis for the Ramanujan zeta function states that all zeros of \(\Xi _R(s)\) are real.
1.2 Zeros of the approximations \(\Xi _F(s)\)
The purpose of this paper is to study the distribution of the zeros of certain approximations for the Ramanujan \(\Xi \)-function. Inspired in the representation (1.1), Ki [2] defined these approximations as follows: Let F be a finite sequence of complex numbers \(a_0, a_1, \ldots , a_n\) such that at least one of them is different from zero. We define the function
where
We recall that \(\overline{\Xi _F(\overline{s})}=\Xi _F(s)\), and one can see that for some sequences \(F_k\), the function \(\Xi _{F_k}(s)\) converges uniformly to \(\Xi _R(s)\) on all compact subsets of \(\mathbb {C}\).
Throughout this paper, we will study the distribution of the zeros of the function \(C_F(s):=\Xi _F(-is)\). Note that the zeros of \(C_F(s)\) are symmetric respect to the line \(\mathrm{Re}\,{s}=0\). Using the argument principle, Ki [2, Theorem 1] established for \(T\ge 2\) thatFootnote 1
where \(N(T,C_F)\) stands for the number of zeros of \(C_F(s)\) such that \(1\le \mathrm{Im}\,{s} < T\), counting multiplicity. In the lower half-plane a similar result holds. Moreover, using the method developed by Levinson [3], he stated that
where \(\overline{N}(T, C_F)\) stands for the number of zeros of \(C_F(s)\) such that \(|\mathrm{Im}\,{s}| < T\), counting multiplicity and \(\overline{N_1}(T,C_F)\) denotes the number of simple zeros such that \(|\mathrm{Im}\,{s}|<T\) and \(\mathrm{Re}\,{s}=0\). In a sense, it means that almost all zeros of \(C_F(s)\) lie on the line \(\mathrm{Re}\,{s}=0\) and are simple. Our first goal is to establish a refinement of (1.2).
Theorem 1
For \(T \ge 2\) we have
On the other hand, Ki [2, Theorem 2] found a result about the vertical distribution of the zeros of \(C_F(s)\), based on the zeros of the function \(\psi _F(s)\), defined by
Let \(k\ge 0\) be an integer such that \(P(1)=P'(1)=\cdot \cdot \cdot =P^{(k-1)}(1)=0\) and \(P^{(k)}(1)\ne 0\), where \(P(y)=\sum _{m=0}^n{a_m}y^{m}\).
Theorem 2
Let \(\Delta _{*}<\Delta _{**}\) be positive real numbers. Suppose that \({\psi _F}(s-k)\) has finitely many zeros in \(-\Delta _{**}<\mathrm{Re}\,{s}<\Delta _{*}\). Let \(\delta \) be such that \(0<\delta <\Delta _{*}\). Then all but finitely many zeros of \({C_F}(s)\) which lie in \(|\mathrm{Re}\,{s}|\le \delta \) are on the line \(\mathrm{Re}\,{s}=0\). In particular, all but finitely many zeros of \({C_F}(s)\) are on the line \(\mathrm{Re}\,{s}=0\), if \({\psi _F}(s-k)\) has finitely many zeros in \(\mathrm{Re}\,{s}>-\Delta _{**}\).
Ki included a second proof for the second part of Theorem 2. In particular, this second proof gave information about the simplicity of the zeros of \(C_F(s)\). Anyway, Ki conjectured that second case for \(\psi _F(s-k)\) is not possible. On the other hand, using (2.7) it is clear that \({\psi _F}(s-k)\) has the same set of zeros of a Dirichlet polynomial in the framework of [4, Sect. 12.5]. The set of zeros of a Dirichlet polynomial is quasi-periodic (see [5, Appendix 6, p. 449]). Then, if \(s_0=\sigma _0+i\tau _0\) is a zero of the Dirichlet polynomial, for any \(\varepsilon >0\) we can construct a sequence \(\{s_n=\sigma _n+i\tau _n\}_{n\in \mathbb {N}}\) of zeros, such that \(\sigma _n\in ]\sigma _0-\varepsilon , \sigma _0+\varepsilon [\) for all \(n\in \mathbb {N}\) and \(\tau _n\rightarrow \pm \infty \). This implies that each open vertical strip has no zeros or has infinite zeros. Therefore, the hypothesis in Theorem 2 is reduced to \({\psi _F}(s-k)\) having no zeros in \(-\Delta _{**}<\mathrm{Re}\,{s}<\Delta _{*}\). Our second goal in this paper is to give a new proof of this result.
Theorem 3
Let \(\Delta _{*}<\Delta _{**}\) be positive real numbers. Suppose that \({\psi _F}(s-k)\) has no zeros in \(-\Delta _{**}<\mathrm{Re}\,{s}<\Delta _{*}\). Let \(\delta \) be such that \(0<\delta <\Delta _{*}\). Then all but finitely many zeros of \({C_F}(s)\) which lie in \(|\mathrm{Re}\,{s}|\le \delta \) are on the line \(\mathrm{Re}\,{s}=0\) and are simple.
We highlight that our proof includes information about the simplicity of the zeros for the first case. The key relation between the functions \(C_F(s)\) and \(\psi _F(s-k)\) is due to de Bruijn [6, p. 225], who showed that
where \(b_m\) are complex numbers and \(b_k\ne 0\).
1.3 Strategy outline
Our approach is motivated by a result of Velásquez [7, Theorem 36], about the distribution of the zeros of a function of the form \(f(s)= h(s)+h^{*}(2a-s)\), where h(s) is a meromorphic functionFootnote 2, and \(a\in \mathbb {R}\). This result can be regarded as a generalization of the necessary condition of stability for the function h(s), in the Hermite–Biehler theorem [5, 21, Part III, Lecture 27]. In our case, using an auxiliary function \(W_F(s)\), we have the representation \(C_F(s)=h(s)+h^*(-s)\), where \(h(s)=W_F(-is-i/2)\). Some estimates of h(s) due to Ki [2, Theorem 2.1] play an important role to establish the necessary growth conditions in [7, Theorem 36]. On the other hand, the strong relation between the zeros of h(s) and \(\psi _F\)(s) (see (2.5)) implies that one study the distribution of zeros of \(\psi _F\)(s), as a set of zeros of a Dirichlet polynomial.
Throughout the paper, we fix a sequence F. For a function f(s) and the parameters \(\sigma _1<\sigma _2\), and \(T_1<T_2\), we denote the counting function
where, in both cases, the counts are with multiplicity, and
where the count is without multiplicity.
2 Preliminary results
In this section we collect several results for our proof. We highlight that in [2, Proposition 2.3], Ki showed that there is a constant \(\beta _0>0\) such that \(C_F(s)\ne 0\), for \(|\mathrm{Re}\,{s}|\ge \beta _0\). This implies that for \(\beta \ge \beta _0\),
Therefore, we can restrict our analysis of the zeros in vertical strips. Now, let us start to find a new representation for \(C_F(s)\). We define the entire function
where
Then, we obtain the following relation:
If we denote by
we rewrite (2.2) as
This representation allows us to use the following result (see [7, Theorem 36]).
Theorem 4
Let \(\sigma _0>0\) be a parameter and h(s) be an entire function such that \(h(s)\ne 0\) for \(\mathrm{Re}\,{s}=\sigma _0\). We define the entire function
Suppose that the function
satisfies the following conditions.
- (i):
-
\(F(s)\ne \pm 1\) on the line \(\mathrm{Re}\,{s}=\sigma _0\), and for some \(\tau _0>0\) we have \(|F(s)|<1\) for \(s=\sigma _0+i\tau \) with \(|\tau |\ge \tau _0\).
- (ii):
-
There exist an increasing function \(\varphi :\mathbb {R}\rightarrow \mathbb {R}\), a constant \(K>0\) and sequences \(\{T_m\}_{m\in \mathbb {N}}\), \(\{T_m^{*}\}_{m\in \mathbb {N}}\) such that \(\displaystyle \lim _{m\rightarrow \infty }T_m=\displaystyle \lim _{m\rightarrow \infty }T_m^{*}=\infty \),
$$\begin{aligned} T_m\le T_{m+1} \le \varphi (T_m), \quad T_m^{*}\le T_{m+1}^{*} \le \varphi (T_m^{*}) \quad \text {for m} \in \mathbb {N}, \end{aligned}$$and \(|F(s)|<e^{K|s|}\), for \(s=\sigma +i\tau \) with \(0\le \sigma \le \sigma _0\) and \(\tau =T_m\), \(\tau =-T_m^{*}\), for \(m\in \mathbb {N}\).
Then, for \(T\ge 2\), we have that
To prove that the function h(s) defined in (2.3) satisfies the conditions of the previous theorem, we will use the estimates used by Ki. By [2, Eq. (2.1)], using the change of variable \(s\mapsto -is-i/2\), we have that
holds uniformly on the half-plane \(\mathrm{Re}\,{s}\ge -1/4\) and |s| sufficiently large. On the other hand, by [2, Theorem 2.1] it follows using the change of variable \(s\mapsto -is+i/2\): for \(\Delta >0\) sufficiently large,
for \(s=\sigma +i\tau \) with \(0\le \sigma \le \Delta \) and \(|\tau |\ge 1\), and the function \(\mu (\sigma )\) is given by
Finally, we will need to establish bounds for the right-hand side of (2.4) to estimate the number of zeros of h(s). The relation (2.5) tells us that we must study the behavior of the zeros of \(\psi _F(s)\). We define \(\psi _{F,k}(s):=\psi _F(s-k)\). Thus, using (1.3) this function can be written as
where \(p_m=(a_{n-m})e^{-\beta _mk}\) and \(\beta _m=\ln ((2n+1)/(2(n-m)+1))\), for \(0\le m\le n\). The sum on the right-hand side of (2.7) is a Dirichlet polynomial in the framework [4, Sect. 12.5].
Proposition 5
Let \(Z(\psi _{F,k})\) denote the set of zeros of \(\psi _{F,k}(s)\).
-
(1)
There is a positive real number \(c_0\) such that \(Z(\psi _{F,k})\subset \{s\in \mathbb {C}: |\mathrm{Re}\,{s}|<c_0\}\).
-
(2)
For \(T_1<T_2\) and \(c\ge c_0\), we have that
$$\begin{aligned} N(-c, c, T_1, T_2, \psi _{F,k})\le n+\dfrac{\ln (2n+1)}{2\pi }(T_2-T_1). \end{aligned}$$ -
(3)
Let \(K\subset \mathbb {C}\) such that \(|\mathrm{Re}\,{s}|\le M\) for \(s\in K\), and some \(M>0\). Suppose that K is uniformly bounded from the zeros of \(\psi _{F,k}(s)\), i.e.,
$$\begin{aligned}\inf \{|s-z|:s\in \textit{K}, z\in Z(\psi _{F,k})\}>0.\end{aligned}$$Then, \(\inf \{|\psi _{F,k}(s)|: s\in \textit{K}\}>0\).
Proof
See [4, Theorems 12.4, 12.5 and 12.6]. \(\square \)
3 Proofs of Theorems 1 and 3
3.1 Proof of Theorem 1
Let us define the function
Since that h(s) and \(h^*(-s)\) are entire functions, we can choose \(\sigma _0>0\) sufficiently large such that \(F(s)\ne \pm 1\) and \(h(s)\ne 0\) on the line \(\mathrm{Re}\,{s}=\sigma _0\). Using (2.5) and (2.6) we get for \(s=\sigma +i\tau \) with \(0\le \sigma \le \sigma _0\) and \(|\tau |\) sufficiently large,
Now, we analyze the behavior of F(s) on the line \(\mathrm{Re}\,{s}=\sigma _0\). Note that \(\mu (\sigma _0)=0\). On another hand, the line \(\mathrm{Re}\,{s}=\sigma _0\) is uniformly bounded from the zeros of \(\psi _{F,k}(s)\). Then, recalling that \(b_k\ne 0\), by Proposition 5 and the triangle inequality we get
for \(s=\sigma _0+i\tau \), with \(|\tau |\) sufficiently large. Inserting this in (3.2), it follows
Therefore, for \(s=\sigma _0+i\tau \) with \(|\tau |\) sufficiently large we conclude that \(|F(s)|<1\). This implies (i) of Theorem 4. Let us prove (ii) of Theorem 4. For each \(m\in \mathbb {Z}\) we consider the rectangle
We divide this rectangle into \(2n+1\) subrectangles \(R_{m,j}\) defined by
for \(j\in \{1,2, \ldots ,2n+1\}\). By Proposition 5 we have that \(N(-\sigma _0, \sigma _0, m, m+1, \psi _{F,k})\le 2n\). So, there exists \(j_0\) such that \(\psi _{F,k}(s)\) does not vanish in \(R_{m,j_0}\). Let us write
Note that \(m<T_m<m+1\). Then, if we define \(\varphi (x)=x+2\), we have that
Let \(\textit{K}=\{s\in \mathbb {C}: -\sigma _0<\mathrm{Re}\,{s}<\sigma _0, ~~\mathrm{Im}\,{s}=T_m, m\in \mathbb {Z}\}\). For any \(s\in \textit{K}\), we have that \(|s-z|\ge 1/{2(2n+1)}\), for all \(z\in Z(\psi _{F,k})\). Then \(\textit{K}\) is uniformly bounded from the zeros of \(\psi _{F,k}(s)\). Using Proposition 5 we see that (3.3) holds for \(s\in K\) with |m| sufficiently large. Therefore, in (3.2) we obtain that for \(s=\sigma +i\tau \) with \(0\le \sigma \le \sigma _0\) and \(\tau =T_m\) (|m| sufficiently large) it follows
Using the fact that \(\mu (\sigma )\le 1\), we conclude that
Now, we choose \(T_m^{*}=-T_{-m}\), for all \(m\in \mathbb {N}\). Thus, we obtain (ii) of Theorem 4. Therefore
To conclude we need to bound \(\widehat{N}(0, \sigma _0, -\varphi (2T), \varphi (2T), h)\). Firstly, we choose \(0<\varepsilon <1/4\) such that h(s) and \(\psi _{F,k}(s)\) do not vanish on \(\mathrm{Re}\,{s}=-\varepsilon _0\). The definition of \(T_m\) implies that
and using Proposition 5 we obtain \(N(-\varepsilon ,\sigma _0,T_m,T_{m+1},\psi _{F,k})\le 2n\). Let us divide the rectangle \(\{s\in \mathbb {C}: -\varepsilon< \mathrm{Re}\,{s}< 0 ~~\text{ and } ~~T_m< \mathrm{Im}\,{s} <T_{m+1}\}\) into \(2n+1\) vertical subrectangles with horizontal length \(\varepsilon /(2n+1)\). So, one of these rectangles, denoted by \(I_m\), has no zeros of \(\psi _{F,k}(s)\) and h(s). Suppose that the right vertical side of \(I_m\) is contained on the line \(\mathrm{Re}\,{s}=-\varepsilon _{m}\) that we can suppose without loss of generality that does not contain a zero of \(\psi _{F,k}(s)\). Now, if we place a circle of radius \(\delta >0\) sufficiently small (for instance \(\delta <1/(2n+1)(16n)\)) we can enclosed the zeros of the rectangle \(J_m=\{s\in \mathbb {C}: -\varepsilon _m< \mathrm{Re}\,{s}< \sigma _0 ~~\text{ and } ~~ T_m< \mathrm{Im}\,{s}< T_{m+1}\}\) in a contour \(C_m\) such that the distance between \(C_m\) and \(J_m\) is at least \(1/(2n+1)(16n)\) and \(C_m\) is distanced at least \(1/(2n+1)(32n)\) from the zeros of \(\psi _{F,k}(s)\). We set \(K=\bigcup _{m} \partial J_m (\partial J_m=\text{ the } \text{ boundaries } \text{ of } J_m).\) Then, applying (2.5) and Proposition 5, we can find \(M>0\) such that
for \(s\in K\), with |m| sufficiently large. If we denote \(w(s)=h(s)/\Gamma (s-k)\), applying Rouché’s theorem we obtain that there is \(m_0\in \mathbb {N}\) sufficiently large such that
and
for \(m\ge m_0\).
Finally, let T be a positive real parameter. If \(T<T_{m_{0}}\) we obtain \(N(0,\sigma _0,0,T,h)=O(1)\). If \(T\ge T_{m_0}\), we choose \(m_1\ge m_0\ge 1\) such that \(m_1<T_{m_1}\le T<T_{m_1+1}<m_1+2\). Since that the zeros of \(1/\Gamma (s)\) are the non-positive integers, by (3.6) , Proposition 5 and (3.5), we get
Similarly, for \(T<0\) we use (3.7) to obtain a similar bound. Thus, we obtain for \(T>0\) that
We replace T by \(\varphi (2T)\) in the above expression, and inserting in (3.4), and one can see that
To obtain our desired result we will use an argument of Ki in [2, p. 131]. Following his idea, for \(T>0\) we get that
where \(\overline{N_k}(T,C_F)\) denotes the number of zeros of \(C_F\) with multiplicity k with \(|\mathrm{Im}\,{s}|< T\) and \(\mathrm{Re}\,{s}=0\), counting with multiplicity. Note that
We conclude combining (3.8), (3.9), (3.10), and recalling by (2.1) that \( \overline{N}(T,C_F)=N(-\sigma _0, \sigma _0, -T ,T , C_F)\).
3.2 Proof of Theorem 3
The proof is similar to the previous case. Using the function defined in (3.1), without loss of generality we can choose \(\delta >0\) in such a way that \(F(s)\ne \pm 1\), \(h(s)\ne 0\) and \(C_F(s)\ne 0\) when \(\sigma =\delta \). By (2.5) and (2.6) it follows for \(s=\sigma +i\tau \) with \(0\le \sigma \le \delta \) and \(|\tau |\) sufficiently large
Using the fact that the \(\psi _{F,k}(s)\) has no zeros in the strip \(-\Delta _{**}<\mathrm{Re}\,{s}<\Delta _{*}\), by Proposition 5 we get
for \(s=\sigma +i\tau \), with \(0\le \sigma \le \delta \) and \(|\tau |\) sufficiently large. Therefore
Using the fact that \(\mu (\delta )<1\), then
for \(s=\delta +i\tau \), with \(|\tau |\) sufficiently large. Further, we have that \(\mu (\sigma )\le 1\), which implies in (3.12) that
for \(s=\sigma +i\tau \) with \(0\le \sigma \le \delta \) and \(|\tau |\) sufficiently large. Choosing \(\varphi (x)=x+2\) and \(T_m=T^*_{m}=m\), for m sufficiently large, we get that the hypotheses in Theorem 4 are satisfied. Then
Combining (2.5) and (3.11), we get a constant \(L>0\) such that \( |h(s)|\ge L|\Gamma (s-k)| \) for \(s=\sigma +i\tau \) with \(0\le \sigma \le \delta \) and \(|\tau |\) sufficiently large. Then, h(s) only has finitely many zeros on the strip \(0\le \sigma \le \delta \), because all possible zeros are contained in a compact set. Therefore, the right-hand side in (3.13) is bounded and this implies our desired result.
Notes
Throughout the paper we use the Vinogradov’s notation \(f=O(g)\) (or \(f\ll g\)) to mean that \(|f(t)|\le C|g(t)|\) for a certain constant \(C>0\) and t sufficiently large.
For a meromorphic function h(s), we define the function \(h^*(s)=\overline{h({\overline{s}})}\).
References
Hardy, G.H.: Ramanujan: Twelve Lectures on Subjects Suggested by His Life and Work. Cambridge University Press, Cambridge (1940)
Ki, H.: On the zeros of approximations of the Ramanujan \(\Xi \)-function. Ramanujan J. 17(1), 123–143 (2008)
Levinson, N.: On theorems of Berlowitz and Berndt. J. Number Theory 3, 502–504 (1971)
Bellman, R., Cooke, K.L.: Differential-Difference Equations. Academic Press, New York (1963)
Levin, B.J.: Distribution of Zeros of Entire Functions. American Mathematical Society, Providence (1956)
de Bruijn, N.G.: The roots of trigonometric integrals. Duke Math. J. 17, 197–226 (1950)
Velásquez, O.: Majoration du nombre de zéros d’une fonction méromorphe en dehors d’une droite verticale et applications. J. Anal. Math. 110, 67–127 (2010)
Acknowledgements
Part of the project was completed during my stay at IMCA with excellent working conditions.
Funding
Open Access funding provided by NTNU Norwegian University of Science and Technology (incl St. Olavs Hospital - Trondheim University Hospital).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
A. C. was supported by Grant 275113 of the Research Council of Norway. O. V. was supported by MATH-AmSud program: project MZFTTA, Peruvian Grant Nos. 310-2014-FONDECYT and CG-176-2015-FONDECYT.
Rights and permissions
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/.
About this article
Cite this article
Chirre, A., Castañón, O.V. A note on the zeros of approximations of the Ramanujan \(\Xi \)-function. Ramanujan J 57, 389–400 (2022). https://doi.org/10.1007/s11139-020-00321-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11139-020-00321-7
Keywords
- Ramanujan zeta function
- Riemann hypothesis
- Zeros of approximations of the Ramanujan \(\Xi \)-function
- Distribution of zeros of entire functions