Equivalent criteria for the Riemann hypothesis for a general class of $L$ -functions

Meghali Garg Meghali Garg
Department of Mathematics
Indian Institute of Technology Indore
Simrol, Indore, Madhya Pradesh 453552, India. [email protected], [email protected] and Bibekananda Maji Bibekananda Maji
Department of Mathematics
Indian Institute of Technology Indore
Simrol, Indore, Madhya Pradesh 453552, India. [email protected], [email protected]

Abstract.

In 1916, Riesz gave an equivalent criterion for the Riemann hypothesis (RH). Inspired from Riesz’s criterion, Hardy and Littlewood showed that RH is equivalent to the following bound:

\displaystyle P_{1}(x):=\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left({-\frac{x% }{n^{2}}}\right)=O_{\epsilon}\left(x^{-\frac{1}{4}+\epsilon}\right),\quad% \mathrm{as}\,\,x\rightarrow\infty.

Recently, the authors extended the above bound for the generalized Riemann hypothesis for Dirichlet $L$ -functions and gave a conjecture for a class of “nice” $L$ -functions. In this paper, we settle this conjecture. In particular, we give equivalent criteria for the Riemann hypothesis for $L$ -functions associated to cusp forms. We also obtain an entirely novel form of equivalent criteria for the Riemann hypothesis of $\zeta(s)$ . Furthermore, we generalize an identity of Ramanujan, Hardy and Littlewood for Chandrasekharan-Narasimhan class of $L$ -functions.

2010 Mathematics Subject Classification. Primary 11M06; Secondary 11M26.
Keywords and phrases. Riemann zeta function,

L

-functions, Non-trivial zeros, Riemann hypothesis, Equivalent criteria.

1. introduction

The Riemann hypothesis, formulated by Bernhard Riemann in his seminal paper [26] in 1859, stands as one of the most renowned yet unsolved problems in the realm of mathematics. Until now, no counterexample has been discovered, and the hypothesis receives substantial numerical support. Nevertheless, despite the persistent endeavours of mathematicians over time, the Riemann hypothesis continues to elude proof. Over the course of time, mathematicians have put forth numerous equivalent criterion for the Riemann hypothesis in their attempts to establish its proof. A noteworthy equivalent criterion, attributed to Riesz [25] in 1916 has shown that the Riemann hypothesis is equivalent to the following bound for an infinite series associated with $\mu(n)$ ,

P_{2}(x):=\sum_{n=1}^{\infty}\frac{\mu(n)}{n^{2}}\exp\left(-\frac{x}{n^{2}}% \right)=O_{\epsilon}\left(x^{-\frac{3}{4}+\epsilon}\right),\quad{\rm as}\,\,x% \rightarrow\infty,

(1.1)

for any positive $\epsilon$ . Driven by this motivation, Hardy and Littlewood [16, p. 161] established another equivalent criterion for the Riemann hypothesis while rectifying an identity of Ramanujan. Mainly, they showed that

\displaystyle P_{1}(x):=\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left({-\frac{x% }{n^{2}}}\right)=O_{\epsilon}\left(x^{-\frac{1}{4}+\epsilon}\right),\quad% \mathrm{as}\,\,x\rightarrow\infty.

(1.2)

Their intuition stemmed from the following identity found in Ramanujan’s notebooks [24, p. 312], [5, Equation (37.3), p. 470], where he mentioned that for any $x>0$ ,

\displaystyle\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left({-\frac{x}{n^{2}}}% \right)=\sqrt{\frac{\pi}{x}}\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left(-% \frac{\pi^{2}}{n^{2}x}\right).

(1.3)

Berndt [5, p. 468-469] presented a compelling numerical explanation that sheds light on the discrepancy of the identity (1.3). Hardy and Littlewood [16, p. 156, Section 2.5] established a corrected version of (1.3). They proved that, for $x>0$ ,

\displaystyle\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left({-\frac{x}{n^{2}}}% \right)=\sqrt{\frac{\pi}{x}}\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left(-% \frac{\pi^{2}}{n^{2}x}\right)-\frac{1}{2\sqrt{\pi}}\sum_{\rho}\left(\frac{\pi}% {\sqrt{x}}\right)^{\rho}\frac{\Gamma\left(\frac{1-\rho}{2}\right)}{\zeta^{% \prime}(\rho)}.

(1.4)

A symmetric form of the above identity is the following. For two positive real numbers $\alpha$ , $\beta$ with $\alpha\beta=\pi$ , we have

\displaystyle\sqrt{\alpha}\sum_{n=1}^{\infty}\frac{\mu(n)}{n}\exp\left({-\left% (\frac{\alpha}{n}\right)^{2}}\right)-\sqrt{\beta}\sum_{n=1}^{\infty}\frac{\mu(% n)}{n}\exp\left({-\left(\frac{\beta}{n}\right)^{2}}\right)=-\frac{1}{2\sqrt{% \beta}}\sum_{\rho}\frac{\Gamma\left(\frac{1-\rho}{2}\right)\beta^{\rho}}{\zeta% ^{\prime}(\rho)},

(1.5)

where the right-side sum over $\rho$ runs through the non-trivial zeros of the Riemann zeta function $\zeta(s)$ . The identity (1.4) holds under the assumption that, all the non-trivial zeros of $\zeta(s)$ are simple. The convergence of the sum over $\rho$ on the right-hand side of (1.4) is not immediately evident. While there is a belief that the series converges rapidly, to date, its convergence is only established under the assumption of bracketing the terms. That is, if for some positive constant $A_{0}$ , the terms corresponding to the non-trivial zeros $\rho_{1}$ and $\rho_{2}$ satisfy

\displaystyle|\operatorname{Im}(\rho_{1})-\operatorname{Im}(\rho_{2})|<\exp% \left(-\frac{A_{0}\operatorname{Im}(\rho_{1})}{\log(\operatorname{Im}(\rho_{1}% ))}\right)+\exp\left(-\frac{A_{0}\operatorname{Im}(\rho_{2})}{\log(% \operatorname{Im}(\rho_{2}))}\right),

(1.6)

then they will be considered inside the same bracket. A more detailed explanation of the convergence intricacies concerning series containing $\zeta^{\prime}(\rho)$ in the denominator has been provided in [19, p. 783].

The identity (1.4) has captured the attention of numerous mathematicians over the years. For further insights into this identity, interested readers are encouraged to refer to Berndt [5, p. 470], Paris and Kaminski [23, p. 143], and Titchmarch [27, p. 219]. Remarkably, Bhaskaran [6] discovered a connection between Fourier reciprocity and Wiener’s Tauberian theory. Ramanujan himself indicated a fascinating generalization of (1.4) related to a pair of reciprocal functions, which was later explored by Hardy and Littlewood [16, p. 160, Section 2.5].

In 2012, Dixit [9] established an elegant character analogue of (1.4). Subsequently, Dixit, Roy, and Zaharescu [12] investigated a one-variable generalization of (1.4), deriving equivalent criteria for both the Riemann hypothesis and the generalized Riemann hypothesis. They [11] also found an identity analogous to (1.4) for $L$ -functions associated with Hecke eigenforms. In $2022$ , Dixit, Gupta, and Vatwani [10] obtained a generalization for the Dedekind zeta function and established an equivalent criterion for the extended Riemann hypothesis. In 2023, Banerjee and Kumar [4] derived an analogous identity for $L$ -functions associated to Maass cusp forms and discovered Riesz-Hardy-Littlewood type equivalent criterion for the corresponding $L$ -function. Very recently, Gupta and Vatwani [15] further examined this identity for $L$ -functions in the Selberg class and obtained Riesz type equivalent criteria. Interested people can also explore the work of Báez-Duarte [3], where a sequential Riesz-type criterion for the Riemann hypothesis was presented. This phenomenon was further studied by Cislo and Wolf [8].

In a recent development, Agarwal and the authors [1] gave an interesting one-variable generalization of the identity of Hardy-Littlewood (1.4). The following identity has been established.

Theorem 1.1.

Let $k\geq 1$ be a real number. Under the hypothesis that the non-trivial zeros of $\zeta(s)$ are simple, for $x>0$ , we have

\displaystyle\sum_{n=1}^{\infty}\frac{\mu(n)}{n^{k}}\exp\left({-\frac{x}{n^{2}% }}\right)=\frac{\Gamma(\frac{k}{2})}{x^{\frac{k}{2}}}\sum_{n=1}^{\infty}\frac{% \mu(n)}{n}{}_{1}F_{1}\left(\frac{k}{2};\frac{1}{2};-\frac{\pi^{2}}{n^{2}x}% \right)+\frac{1}{2}\sum_{\rho}\frac{\Gamma(\frac{k-\rho}{2})}{\zeta^{\prime}(% \rho)}x^{-\frac{(k-\rho)}{2}},

(1.7)

where the sum over $\rho$ runs through all the non-trivial zeros of $\zeta(s)$ and satisfies the bracketing condition (1.6) and the entire function ${}_{1}F_{1}(a;b;z)$ is defined by

\displaystyle{}_{1}F_{1}(a;b;z):=\sum_{n=0}^{\infty}\frac{(a)_{n}}{(b)_{n}}% \frac{z^{n}}{n!},\quad{where}~{}~{}(a)_{n}:=\frac{\Gamma(a+n)}{\Gamma(a)}.

Building upon Dixit’s [9] work, the authors [14] provided a character analogue of (1.7) and established the identity as follows:

Theorem 1.2.

Let $k\geq 1$ and $\chi$ be a primitive character modulo $q$ . Assume that all the non-trivial zeros of $L(s,\chi)$ are simple. For $x>0$ , we have

$\displaystyle\sum_{n=1}^{\infty}\frac{\chi(n)\mu(n)}{n^{k}}\exp\left({-\frac{% \pi x^{2}}{qn^{2}}}\right)$	$\displaystyle=\frac{i^{a}\sqrt{q}}{G(\chi)}\left(\frac{q}{\pi x^{2}}\right)^{% \frac{k+a}{2}}\left(\frac{\pi}{q}\right)^{a+\frac{1}{2}}\frac{\Gamma(\frac{k+a% }{2})}{\Gamma(a+\frac{1}{2})}$
	$\displaystyle\times\sum_{n=1}^{\infty}\frac{\overline{\chi(n)}\mu(n)}{n^{1+a}}% {}_{1}F_{1}\left(\frac{k+a}{2};a+\frac{1}{2};-\frac{\pi}{qn^{2}x^{2}}\right)$
	$\displaystyle+\frac{1}{2}\sum_{\rho}\frac{\Gamma(\frac{k-\rho}{2})}{L^{\prime}% (\rho,\chi)}\left(\frac{q}{\pi x^{2}}\right)^{\frac{k-\rho}{2}},$	(1.8)

where the sum over $\rho$ runs through the non-trivial zeros of $L(s,\chi)$ and satisfy the bracketing conditions (1.6).

Inspired by the works of Riesz, Hardy and Littlewood, the authors [14] have established the following equivalent criteria for the generalized Riemann hypothesis for $L(s,\chi)$ . For $k\geq 1$ and $\ell>0$ , we have

\sum_{n=1}^{\infty}\frac{\chi(n)\mu(n)}{n^{k}}\exp\left(-\frac{x}{n^{\ell}}% \right)=O_{\epsilon,k,\ell}\left(x^{-\frac{k}{\ell}+\frac{1}{2\ell}+\epsilon}% \right),\quad\mathrm{as}\,\,x\rightarrow\infty,

(1.9)

for any $\epsilon>0$ . This bound extends the previously established bounds by Riesz (1.1), Hardy-Littlewood (1.2), as well as the earlier bound given by Agarwal and the authors [1, Theorem 1.2]. This bound, in turn, has motivated us to propose a broader conjecture [14, Conjecture 5.1], which we restate below for the clarity of the readers. Let us consider

\displaystyle L(f,s)=\sum_{n=1}^{\infty}\frac{A_{f}(n)}{n^{s}},~{}{\rm{for}}~{% }\operatorname{Re}(s)>1,

(1.10)

be a “nice” $L$ -function. Mainly, we are assuming that $L(f,s)$ obeys the grand Riemann hypothesis [18, p. 113], which says that all the non-trivial zeros of $L(f,s)$ in the critical strip $0<\operatorname{Re}(s)<1$ lie on the critical line $\operatorname{Re}(s)=\frac{1}{2}$ . Let us suppose that

\displaystyle 1/L(f,s)=\sum_{n=1}^{\infty}\frac{\mu_{f}(n)}{n^{s}},~{}\text{% for}~{}\operatorname{Re}(s)>1.

(1.11)

Moreover, we assume that $\sum_{n=1}^{\infty}\mu_{f}(n)n^{-1}$ is convergent. Littlewood’s prediction says that the grand Riemann hypothesis [18, Proposition 5.14] is equivalent to the following bound:

\displaystyle\sum_{n\leq x}\mu_{f}(n)\ll_{\epsilon,f}x^{\frac{1}{2}+\epsilon},

(1.12)

for any $\epsilon>0$ . The above bound played a significant role to prove (1.9). Inspired from this observation, the authors gave the following conjecture.

Conjecture 1.3.

Let $k\geq 1$ and $\ell>0$ be two real numbers. Let $L(f,s)$ be a “nice” $L$ -function satisfying $\eqref{merten bound general}$ . Then the grand Riemann hypothesis for $L(f,s)$ is equivalent to the bound

\displaystyle P_{k,f,\ell}:=\sum_{n=1}^{\infty}\frac{\mu_{f}(n)}{n^{k}}\exp% \left(-\frac{x}{n^{\ell}}\right)=O_{\epsilon,k,\ell}\left(x^{-\frac{k}{\ell}+% \frac{1}{2\ell}+\epsilon}\right),\quad\mathrm{as}\,\,x\rightarrow\infty.

(1.13)

In this manuscript, one of our main goals is to prove the above conjecture. Moreover, we also establish a generalization of the Ramanujan-Hardy-Littlewood identity (1.4) within the framework of Chandrasekharan-Narasimhan class of $L$ -functions [7], defined in the next section. Furthermore, we obtain an entirely novel form of equivalent criteria for the Riemann hypothesis of $\zeta(s)$ , see Theorem 3.13.

2. Chandrasekharan-Narasimhan class of $L$ -functions

We now introduce the class of $L$ -functions established by Chandrasekharan and Narasimhan [7]. Consider two arithmetical functions, denoted as $a_{1}(n)$ and $b_{1}(n)$ , which are not identically zero. Additionally, let ${\lambda_{n}}$ and ${\mu_{n}}$ be two monotonically increasing sequences of positive real numbers tending to infinity. Then the associated Dirichlet series are defined as follows:

	$\displaystyle\phi(s)=\sum_{n=1}^{\infty}\frac{a_{1}(n)}{\lambda_{n}^{s}}\quad{% \rm{for}}\quad\operatorname{Re}(s)=\sigma>\sigma_{a},$		(2.1)
	$\displaystyle\psi(s)=\sum_{n=1}^{\infty}\frac{b_{1}(n)}{\mu_{n}^{s}}\quad{\rm{% for}}\quad\operatorname{Re}(s)=\sigma>\sigma_{b}.$		(2.2)

Here, $\sigma_{a}$ and $\sigma_{b}$ indicate the finite abscissae of absolute convergence for $\phi(s)$ and $\psi(s)$ , respectively. It is assumed that both $\phi(s)$ and $\psi(s)$ can be analytically continued to the entire $\mathbb{C}$ except for a finite number of poles. For a given $\delta>0$ , we assume that $\phi(s)$ and $\psi(s)$ are connected by the subsequent functional equation:

\displaystyle(2\pi)^{-s}\Gamma(s)\phi(s)=(2\pi)^{-(\delta-s)}\Gamma(\delta-s)% \psi(\delta-s).

(2.3)

In this paper, we adopt their framework in a slightly general form stated below. Essentially, we consider a definition that involves a certain twist. We define

\displaystyle a(n)=\begin{cases}a_{1}(k),&\textit{if}\quad n=\lambda_{k},\\ 0,&otherwise,\end{cases}\quad\text{and}\quad b(n)=\begin{cases}b_{1}(k),&% \textit{if}\quad n=\mu_{k},\\ 0,&otherwise.\end{cases}

Thus, the Dirichlet series in (2.1) and (2.2) can be rewritten as follows:

	$\displaystyle\phi(s)=\sum_{n=1}^{\infty}\frac{a(n)}{n^{s}},\quad\text{for}% \quad\operatorname{Re}(s)=\sigma>\sigma_{a},$
	$\displaystyle\psi(s)=\sum_{n=1}^{\infty}\frac{b(n)}{n^{s}},\quad\text{for}% \quad\operatorname{Re}(s)=\sigma>\sigma_{b}.$

Now, for a specified $\delta>0$ , we assume that $\phi(s)$ and $\psi(s)$ satisfy the following general functional equation:

\displaystyle c^{s}\Gamma(As+B)\phi(s)=\nu c^{\delta-s}\Gamma(A(\delta-s)+B)% \psi(\delta-s),

(2.4)

where $|\nu|=1$ , $A>0$ , $B\geq 0$ and $c\in\mathbb{R}^{+}$ . One can easily see that (2.3) is a particular case of (2.4) by letting $c=\frac{1}{2\pi},A=1,B=0$ and $\nu=1.$

We can observe that $\zeta(s)$ and $L(s,\chi)$ trivially fall under the category of the above class of $L$ -functions. A few more examples that belong to the Chandrasekharan-Narasimhan class of $L$ -functions are given below.

Example 1.

Let $r\in\mathbb{Z}$ and $\sigma_{r}(n)$ denote the sum of $r^{th}$ powers of positive divisors of $n$ . Then the $L$ -function associated to $\sigma_{r}(n)$ is given by

\displaystyle\ \sum_{n=1}^{\infty}\frac{\sigma_{r}(n)}{n^{s}}=\zeta(s)\zeta(s-% r),\quad{\rm{for}}\quad\operatorname{Re}(s)=\sigma>\max\{1,r+1\},

and if $r$ is an odd positive integer, it satisfies the following functional equation:

\displaystyle(2\pi)^{-s}\Gamma(s)\zeta(s)\zeta(s-r)=(-1)^{\frac{r+1}{2}}(2\pi)% ^{-(r+1-s)}\Gamma(r+1-s)\zeta(r+1-s)\zeta(1-s).

(2.5)

This corresponds to (2.4) for $a(n)=b(n)=\sigma_{r}(n),c=\frac{1}{2\pi},A=1,B=0,\delta=r+1$ and $\nu=(-1)^{\frac{r+1}{2}}.$ One can refer [7, Example 2] for further insight into the functional equation (2.5).

An intriguing observation about Example 1 is that, despite the function $\zeta(s)\zeta(s-r)$ having non-trivial zeros on the lines $\operatorname{Re}(s)=\frac{1}{2}$ and $\operatorname{Re}(s)=r+\frac{1}{2}$ , and thus not adhering its own Riemann hypothesis but it offers surprising insights into the Riemann hypothesis for $\zeta(s)$ by presenting an entirely novel form of equivalent criteria, as demonstrated in Theorem 3.13. The next example corresponds to the cusp form of weight $12$ .

Example 2.

The Ramanujan tau function $\tau(n)$ is defined by

\displaystyle\Delta(q)=\sum_{n=1}^{\infty}\tau(n)q^{n}=q\prod_{n=1}^{\infty}(1% -q^{n})^{24},\quad|q|<1.

Note that $\Delta(q)$ is a cusp form of level $1$ and weight $12$ . The $L$ -function associated to normalized coefficients $\tau_{0}(n)=\tau(n)n^{-\frac{11}{2}}$ is given by

\displaystyle L(\Delta,s)=\sum_{n=1}^{\infty}\frac{\tau_{0}(n)}{n^{s}}\quad{% \rm{for}}\quad\operatorname{Re}(s)=\sigma>1,

and can be analytically continued to the whole complex plane as an entire function. Moreover, $L(\Delta,s)$ satisfies the following nice-looking functional equation:

\displaystyle(2\pi)^{-s}\Gamma\left(s+\frac{11}{2}\right)L(\Delta,s)=(2\pi)^{-% (1-s)}\Gamma\left(\frac{13}{2}-s\right)L(\Delta,1-s),

which aligns with (2.4) for $a(n)=b(n)=\tau_{0}(n),c=\frac{1}{2\pi},A=1,B=\frac{11}{2}$ , $\delta=1$ and $\nu=1$ . It has been conjectured that all the non-trivial zeros of $L(\Delta,s)$ lie on the critical line $\operatorname{Re}(s)=\frac{1}{2}$ .

More general setting for $L$ -functions associated to cusp forms are stated below.

Example 3.

Let $f(z)$ be a holomorphic Hecke eigenform of weight $\omega$ for the full modular group $SL(2,\mathbb{Z})$ . The Fourier expansion of $f(z)$ at $i\infty$ is given by

\displaystyle f(z)=\sum_{n=1}^{\infty}\lambda_{f}(n)n^{\frac{\omega-1}{2}}e^{2% \pi inz},

and $\lambda_{f}(n)$ is the normalized $n$ th Fourier coefficient. Then the $L$ -function associated to $f(z)$ is defined by

\displaystyle L(f,s)=\sum_{n=1}^{\infty}\frac{\lambda_{f}(n)}{n^{s}}=\prod_{p:% ~{}\rm{prime}}\left(1-\frac{\lambda_{f}(p)}{p^{s}}+\frac{1}{p^{2s}}\right)^{-1% }\quad{\rm{for}}\quad\operatorname{Re}(s)>1.

(2.6)

Hecke proved that $L(f,s)$ satisfies the functional equation

\displaystyle\Lambda(f,s)=\Lambda(f,1-s),

(2.7)

where $\Lambda(f,s):=(2\pi)^{-s}\Gamma\big{(}s+\frac{\omega-1}{2}\big{)}L(f,s)$ . The $L$ -function $L(f,s)$ admits analytic continuation to the whole complex plane. Here, (2.7) synchronizes with (2.4) for $a(n)=b(n)=\lambda_{f}(n),c=\frac{1}{2\pi},A=1,B=\frac{\omega-1}{2}$ , $\delta=1$ and $\nu=1$ . The Riemann hypothesis for $L(f,s)$ states that all the non-trivial zeros of $L(f,s)$ are situated on the critical line $\operatorname{Re}(s)=\frac{1}{2}$ .

Example 4.

Let $r_{j}(n)$ counts the number of ways a positive integer $n$ can be written as sum of $j$ squares. Then the generating function

\displaystyle\zeta_{id}(j,s):=\sum_{n=1}^{\infty}\frac{r_{j}(n)}{n^{s}}\quad{% \rm{for}}\quad\operatorname{Re}(s)>\frac{j}{2},

is the Epstein zeta function and satisfies the functional equation

\displaystyle\pi^{-s}\Gamma(s)\zeta_{id}(j,s)=\pi^{s-\frac{j}{2}}\Gamma\left(% \frac{j}{2}-s\right)\zeta_{id}\left(j,\frac{j}{2}-s\right),

which is due to Epstein [13]. It can be analytically continued to the whole complex plane except for a simple pole at $s=\frac{j}{2}$ . This corresponds to (2.4) for $a(n)=b(n)=r_{j}(n),c=\frac{1}{\pi},A=1,B=0,\delta=\frac{j}{2}$ and $\nu=1.$ When $j=2$ , it has been conjectured that all the non-trivial zeros of $\zeta_{id}(2,s)$ lies on the critical line $\operatorname{Re}(s)=\frac{1}{2}.$

Exploring whether a given subset $\mathcal{A}$ of natural numbers $\mathbb{N}$ can be used to express any natural number $n$ as a sum of elements from $\mathcal{A}$ , and if so, quantifying the various ways this can be achieved, is always a compelling topic of study. In the context of Example 4, the subset $\mathcal{A}$ is defined as the set of all perfect squares. Typically, Epstein zeta functions are investigated for positive definite $j\times j$ matrix $M$ and are defined as

\displaystyle\zeta_{M}(j,s)=\sum_{\kappa\in\mathbb{Z}^{j}/\{0\}}(\kappa^{T}M% \kappa)^{-s}\quad{\rm{for}}\quad\operatorname{Re}(s)>\frac{j}{2}.

Example 4 deals with $M$ being identity matrix. For $j=2,4,6$ and $8$ , the Epstein zeta function can be expressed in terms of Riemann zeta function and Dirichlet $L$ -functions. An elegant formula for $\zeta_{id}(2,s)$ can be found in [29, (1.1)], which states that

\displaystyle\zeta_{id}(2,s)=4\zeta(s)\beta(s),

(2.8)

where $\beta(s)=\sum_{k=1}^{\infty}\frac{(-1)^{k-1}}{(2k-1)^{s}}$ , which is noting but $L(s,\chi_{4,2})$ . The generalized Riemann hypothesis for $L(s,\chi_{4,2})$ says that all the non-trivial zeros will lie on $\operatorname{Re}(s)=1/2$ . Thus, all the non-trivial zeros of $\zeta_{id}(2,s)$ will also lie on $\operatorname{Re}(s)=\frac{1}{2}$ . However, it has been shown that Epstein zeta functions $\zeta_{id}(j,s)$ , for $j\geq 3$ , do not satisfy the Riemann hypothesis. A detailed explanation for $j=4$ can be found in [28, (1.4)-(1.6)] and numerical evidences from [28, Section 5] ensures the failure of Epstein Riemann hypothesis for higher dimensions.
The next example is a particular case of the well known Dedekind zeta function for the imaginary quadratic field $\mathbb{Q}(\sqrt{-D})$ .

Example 5.

Let $K=\mathbb{Q}(\sqrt{-D})$ be an imaginary quadratic field with discriminant $D$ . Then the Dedekind zeta function is defined as

\displaystyle\zeta_{K}(s):=\sum_{n=1}^{\infty}\frac{\mathsf{a}_{n}}{n^{s}}% \quad{\rm{for}}\quad\operatorname{Re}(s)>1,

(2.9)

where $\mathsf{a}_{n}$ denotes the number of integral ideals of norm $n$ of an imaginary quadratic field $\mathbb{Q}(\sqrt{-D})$ . It satisfies the following functional equation

\displaystyle\left(\frac{2\pi}{\sqrt{D}}\right)^{-s}\Gamma(s)\zeta_{K}(s)=% \left(\frac{2\pi}{\sqrt{D}}\right)^{-(1-s)}\Gamma(1-s)\zeta_{K}(1-s),

(2.10)

due to [20, p. 266]. This fits in (2.4) for $c=\frac{\sqrt{D}}{2\pi},A=1,B=0,\delta=1$ and $\nu=1$ . The extended Riemann hypothesis asserts that all the non-trivial zeros of $\zeta_{K}(s)$ lie on $\operatorname{Re}(s)=\frac{1}{2}$ .

Before stating our main theorems, we need to define the Dirichlet inverse of an arithmetical function $a(n)$ . Let $a(n)$ be an arithmetical function with $a(1)\neq 0$ , then there exist a unique arithmetical function $a^{-1}(n)$ , known as Dirichlet inverse of $a(n)$ , such that

\displaystyle a*a^{-1}=a^{-1}*a=I.

Furthermore, $a^{-1}(n)$ is recursively defined as

\displaystyle a^{-1}(1)=\frac{1}{a(1)},\hskip 14.22636pta^{-1}(n)=\frac{-1}{a(% 1)}\sum_{\begin{subarray}{c}d|n\\ d<n\end{subarray}}a\left(\frac{n}{d}\right)a^{-1}(d)\quad{\rm{for}}\quad n>1.

A proof of this standard result can be found in [2, p. 30]. Now we are ready to state our main results.

3. Main Results

In this section, we state the main results of this paper.

Theorem 3.1.

Let $a(n)$ and $b(n)$ be two arithmetical functions whose corresponding $L$ -functions are $\phi(s)$ and $\psi(s)$ and satisfy the functional equation (2.4). Assume that all the non-trivial zeros of $\phi(s)$ are simple. Then for any $k\geq\delta+\frac{B}{A}$ and $x>0$ , we have

		$\displaystyle\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k}}\exp\left({-\frac{x}{n^% {1/A}}}\right)$
		$\displaystyle=\frac{c^{-\frac{A\delta+2B}{A}}}{\nu x^{Ak+B}}\frac{\Gamma(Ak+B)% }{\Gamma(A\delta+2B)}\sum_{n=1}^{\infty}\frac{b^{-1}(n)}{n^{\frac{A\delta+B}{A% }}}\bigg{[}{}_{1}F_{1}\left(\begin{matrix}Ak+B\\ A\delta+2B\end{matrix}\Big{\|}-\frac{1}{(c^{2}n)^{1/A}x}\right)-1\bigg{]}$
		$\displaystyle-R_{t}+A\sum_{\rho}\frac{\Gamma(A(k-\rho))}{\phi^{\prime}(\rho)}x% ^{-A(k-\rho)},$		(3.1)

where $\rho$ runs through the non-trivial zeros of $\phi(s)$ and the convergence of the infinite series over $\rho$ adhere to the bracketing condition (1.6), and $R_{t}$ represents the residual term corresponding to the trivial zeros of $\phi(k-\frac{s}{A})$ with $0\leq\operatorname{Re}\left(k-\frac{s}{A}\right)\leq Ak+B$ .

Remark 1.

Setting $a(n)=b(n)=1$ and thus $a^{-1}(n)=b^{-1}(n)=\mu(n)$ , $\phi(s)=\psi(s)=\zeta(s)$ in Theorem 3.1 with $A=\frac{1}{2},B=0,c=\frac{1}{\sqrt{\pi}},\delta=1$ and $\nu=1,$ one can easily obtain the identity (1.7).

Remark 2.

Let $\chi$ be a primitive Dirichlet character modulo $q$ and $L(s,\chi)$ be the associated Dirichlet series. Letting $a(n)=\chi(n),b(n)=\overline{\chi(n)}$ and thus $a^{-1}(n)=\chi(n)\mu(n),b^{-1}(n)=\overline{\chi(n)}\mu(n)$ , $\phi(s)=L(s,\chi),\psi(s)=L(s,\overline{\chi})$ in Theorem 3.1 with $A=\frac{1}{2},B=\frac{a}{2},c=\sqrt{\frac{q}{\pi}},\delta=1$ and $\nu=\frac{G(\chi)}{i^{a}\sqrt{q}},$ and upon simplification, we can derive (1.8).

Now we shall state a few more interesting results coming from our Theorem 3.1. First, we state an ingenious modular relation, which is a perfect generalization of Ramanujan-Hardy-Littlewood identity (1.5).

Corollary 3.2.

Assuming all the hypothesis of the Theorem 3.1 to be true. Let $\alpha$ , $\beta$ be positive real numbers such that $\alpha\beta=\frac{1}{c^{2/A}}$ . Then we have

	$\displaystyle\sqrt{\bar{\nu}}\alpha^{\frac{A\delta+2B}{2}}\sum_{n=1}^{\infty}% \frac{b^{-1}(n)}{n^{\delta+B/A}}\exp\left(-\frac{\alpha}{n^{1/A}}\right)-\sqrt% {\nu}\beta^{\frac{A\delta+2B}{2}}\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{\delta% +B/A}}\exp\left(-\frac{\beta}{n^{1/A}}\right)$
	$\displaystyle=\frac{\alpha^{\frac{A\delta+2B}{2}}}{\sqrt{\nu}}\sum_{n=1}^{% \infty}\frac{b^{-1}(n)}{n^{\delta+B/A}}+\sqrt{\nu}\beta^{\frac{A\delta+2B}{2}}% R_{t}-A\sqrt{\nu}\sum_{\rho}\frac{\Gamma(A(\delta-\rho)+B)}{\phi^{\prime}(\rho% )}\beta^{\frac{A\delta-2B}{2}}.$

The next result is corresponding to the Epstein zeta function, defined in Example 4.

Corollary 3.3.

Let $k\geq 1$ and $x$ be any positive real number. Let $r_{2}(n)$ be the number of ways $n$ can be written as sum of two squares and the associated $L$ -function is defined in Example 4 for $j=2.$ Then, we have

\displaystyle\sum_{n=1}^{\infty}\frac{r_{2}^{-1}(n)}{n^{k}}\exp\left(-\frac{x}% {n}\right)=\frac{\pi\Gamma(k)}{x^{k}}\sum_{n=1}^{\infty}\frac{r_{2}^{-1}(n)}{n% }{}_{1}F_{1}\left(\begin{matrix}k\\ 1\end{matrix}\Big{|}-\frac{\pi^{2}}{nx}\right)+\sum_{\rho}\frac{\Gamma(k-\rho)% }{\zeta_{id}^{{}^{\prime}}(2,\rho)}x^{-(k-\rho)}.

(3.2)

Here, $\rho$ runs over all the non-trivial zeros of $\zeta_{id}(2,s)$ and are assumed to be simple. The convergence of the series over $\rho$ follows under the assumption of the bracketing condition (1.6).

A few numerical evidences for the above result is given in Table 1.

Remark 3.

In particular, letting $k=1$ and $x=\beta>0$ with $\alpha\beta=\pi^{2}$ in Corollary 3.3, we obtain the following beautiful identity,

\displaystyle\sqrt{\alpha}\sum_{n=1}^{\infty}\frac{r_{2}^{-1}(n)}{n}\exp\left(% -\frac{\alpha}{n}\right)

\displaystyle-\sqrt{\beta}\sum_{n=1}^{\infty}\frac{r_{2}^{-1}(n)}{n}\exp\left(% -\frac{\beta}{n}\right)=-\frac{1}{\sqrt{\beta}}\sum_{\rho}\frac{\Gamma(1-\rho)% }{\zeta_{id}^{{}^{\prime}}(2,\rho)}\beta^{\rho}.

The next identity is associated to the Dedekind zeta function over an imaginary quadratic field.

Corollary 3.4.

Let $k\geq 1$ be any real number. If $\mathsf{a}_{n}$ counts the number of integral ideals of norm $n$ of an imaginary quadratic field $\mathbb{Q}(\sqrt{-D})$ and the associated $L$ -function is defined as in (2.9) and satisfies the functional equation (2.10), then for any $x>0$ , we have

\displaystyle\sum_{n=1}^{\infty}\frac{\mathsf{a}_{n}^{-1}}{n^{k}}\exp\left(-% \frac{x}{n}\right)

\displaystyle=\frac{2\pi\Gamma(k)}{\sqrt{D}x^{k}}\sum_{n=1}^{\infty}\frac{% \mathsf{a}_{n}^{-1}}{n}{}_{1}F_{1}\left(\begin{matrix}k\\ 1\end{matrix}\Big{|}-\frac{4\pi^{2}}{Dnx}\right)+\sum_{\rho}\frac{\Gamma(k-% \rho)}{\zeta_{K}^{{}^{\prime}}(\rho)}x^{-(k-\rho)},

where $\rho$ runs over the non-trivial zeros of $\zeta_{K}(s)$ and convergence of the infinite series over $\rho$ follows from the bracketing condition (1.6).

Remark 4.

In particular, substituting $k=1$ and $x=\beta>0$ and $\alpha\beta=\frac{4\pi^{2}}{D}$ in Corollary 3.4 gives the following identity¹¹1In [10, Cor. 3.4], there is an extra $1/2$ on the right side of the identity. of Dixit, Gupta and Vatwani [10, Cor. 3.4],

\displaystyle\sqrt{\alpha}\sum_{n=1}^{\infty}\frac{\mathsf{a}_{n}^{-1}}{n}\exp% \left(-\frac{\alpha}{n}\right)-\sqrt{\beta}\sum_{n=1}^{\infty}\frac{\mathsf{a}% _{n}^{-1}}{n}\exp\left(-\frac{\beta}{n}\right)=-\frac{1}{\sqrt{\beta}}\sum_{% \rho}\frac{\Gamma(1-\rho)}{\zeta_{K}^{{}^{\prime}}(\rho)}\beta^{\rho}.

The next result corresponds to $\phi(s)=\zeta(s)\zeta(s-r)$ , defined in Example 1, an important example of a Dirichlet series lying in Chadrasekharan-Narasimhan class of $L$ -functions, which lacks the Riemann hypothesis.

Theorem 3.5.

Let $\sigma_{r}(n)$ denote the sum of $r^{th}$ powers of positive divisors of $n$ . Then the corresponding $L$ -function as defined in Example 1 satisfies the functional equation (2.5). Let $k\geq r+1$ , where $r$ is an odd positive integer. Then for any positive $x$ , we have

	$\displaystyle\sum_{n=1}^{\infty}\frac{\sigma_{r}^{-1}(n)}{n^{k}}\exp\left(-% \frac{x}{n}\right)$	$\displaystyle=\frac{(-1)^{\frac{r+1}{2}}(2\pi)^{r+1}}{x^{k}}\frac{\Gamma(k)}{% \Gamma(r+1)}\sum_{n=1}^{\infty}\frac{\sigma_{r}^{-1}(n)}{n^{r+1}}{}_{1}F_{1}% \left(\begin{matrix}k\\ r+1\end{matrix}\Big{\|}-\frac{4\pi^{2}}{nx}\right)$
		$\displaystyle+\sum_{m=1}^{[\frac{r}{2}]}\frac{\Gamma(k-r+2m)}{\zeta(r-2m)\zeta% (2m+1)}\frac{(-1)^{m}2(2\pi)^{2m}}{(2m)!}x^{r-k-2m}$
		$\displaystyle+\sum_{\rho}\frac{x^{\rho-k}}{\zeta^{{}^{\prime}}(\rho)}\bigg{[}% \frac{\Gamma(k-\rho)}{\zeta(\rho-r)}+\frac{x^{r}\Gamma(k-r-\rho)}{\zeta(r+\rho% )}\bigg{]},$

where $\rho$ runs over non-trivial zeros of $\zeta(s)$ and are assumed to be simple. The convergence of the associated series follows under the assumption of bracketing condition (1.6).

A numerical verification of the identity is given in Table 2.

In particular, letting $k=r+1$ and $\alpha\beta=4\pi^{2}$ in Theorem 3.5, we get the following identity.

Corollary 3.6.

Let $r$ be an odd positive integer and $\alpha,\beta$ be positive real numbers with $\alpha\beta=4\pi^{2}$ . Then we have

	$\displaystyle\alpha^{\frac{r+1}{2}}$	$\displaystyle\sum_{n=1}^{\infty}\frac{\sigma_{r}^{-1}(n)}{n^{r+1}}\exp\left(-% \frac{\alpha}{n}\right)-(-\beta)^{\frac{r+1}{2}}\sum_{n=1}^{\infty}\frac{% \sigma_{r}^{-1}(n)}{n^{r+1}}\exp\left(-\frac{\beta}{n}\right)$
		$\displaystyle=2\alpha^{\frac{r-1}{2}}\sum_{m=1}^{[\frac{r}{2}]}\frac{(-1)^{m}}% {\zeta(r-2m)\zeta(2m+1)}\left(\frac{\beta}{2\pi}\right)^{2m}$
		$\displaystyle+\alpha^{\frac{r+1}{2}-k}\sum_{\rho}\frac{\alpha^{\rho}}{\zeta^{{% }^{\prime}}(\rho)}\bigg{[}\frac{\Gamma(r+1-\rho)}{\zeta(\rho-r)}+\frac{\alpha^% {r}\Gamma(1-\rho)}{\zeta(r+\rho)}\bigg{]}.$

The next result, for $L$ -functions associated to cusp forms, we state as a separate theorem as it might have a significant importance in the literature.

Theorem 3.7.

Let $L(f,s)$ be the $L$ -function associated to a cusp form $f$ defined in (2.6). If we assume all the non-trivial zeros of $L(f,s)$ to be simple, then for $k\geq\frac{w+1}{2}$ , $x>0$ , we have

	$\displaystyle\sum_{n=1}^{\infty}\frac{\lambda_{f}^{-1}(n)}{n^{k}}\exp\left(-% \frac{x}{n}\right)$	$\displaystyle=\frac{(2\pi)^{\omega}}{x^{k+\frac{\omega-1}{2}}}\frac{\Gamma(k+% \frac{\omega-1}{2})}{\Gamma(\omega)}\sum_{n=1}^{\infty}\frac{\lambda_{f}^{-1}(% n)}{n^{\frac{\omega+1}{2}}}{}_{1}F_{1}\left(\begin{matrix}k+\frac{\omega-1}{2}% \\ \omega\end{matrix}\Big{\|}-\frac{4\pi^{2}}{nx}\right)$
		$\displaystyle+\sum_{\rho}\frac{\Gamma(k-\rho)x^{-(k-\rho)}}{L^{{}^{\prime}}(f,% \rho)},$		(3.3)

where $\rho$ runs through non-trivial zeros of $L(f,s)$ and satisfy the bracketing condition (1.6).

By substituting $k=\frac{\omega+1}{2}$ in Theorem 3.7, we obtain an elegant modular relation depicted as follows.

Corollary 3.8.

Assuming all the hypothesis in Theorem 3.7, for $\alpha,\beta>0$ with $\alpha\beta=4\pi^{2}$ , we have

\displaystyle\alpha^{\frac{\omega}{2}}\sum_{n=1}^{\infty}\frac{\lambda_{f}^{-1% }(n)}{n^{\frac{\omega+1}{2}}}\exp\left(-\frac{\alpha}{n}\right)

\displaystyle-\beta^{\frac{\omega}{2}}\sum_{n=1}^{\infty}\frac{\lambda_{f}^{-1% }(n)}{n^{\frac{\omega+1}{2}}}\exp\left(-\frac{\beta}{n}\right)=-\frac{1}{\sqrt% {\beta}}\sum_{\rho}\frac{\Gamma(\frac{\omega+1}{2}-\rho)\beta^{\rho}}{L^{{}^{% \prime}}(f,\rho)}.

The next corollary gives a modular relation involving Ramanujan tau function.

Corollary 3.9.

Let $\alpha$ and $\beta$ be two positive real numbers with $\alpha\beta=4\pi^{2}$ . Then we have

\displaystyle\alpha^{6}\sum_{n=1}^{\infty}\frac{\tau_{0}^{-1}(n)}{n^{\frac{13}% {2}}}\exp\left(-\frac{\alpha}{n}\right)

\displaystyle-\beta^{6}\sum_{n=1}^{\infty}\frac{\tau_{0}^{-1}(n)}{n^{\frac{13}% {2}}}\exp\left(-\frac{\beta}{n}\right)=-\frac{1}{\sqrt{\beta}}\sum_{\rho}\frac% {\Gamma(\frac{13}{2}-\rho)\beta^{\rho}}{L^{{}^{\prime}}(\Delta,\rho)},

(3.4)

where $\rho$ runs through non-trivial zeros of $L(\Delta,s)$ .

It is interesting to note that the above identity has already been obtained by Dixit, Roy and Zaharescu [11, Corollary 1.2]. One can also obtain Corollary 3.8 by letting $z\rightarrow 0$ in Theorem 1.1 of [11]. However, our Theorem 3.7 is different from Theorem 1.1 of Dixit et. al [11].

3.1. Equivalent criteria for the grand Riemann hypothesis

Motivated from the equivalent criteria for the Riemann hypothesis of $\zeta(s)$ due to Hardy-Littlewood and Riesz, in this subsection, we state equivalent criteria for the Riemann hypothesis for a general class of $L$ -functions, defined in (1.10) and satisfying Littlewood’s bound (1.12). In particular, we give equivalent criteria for the Riemann hypothesis for $L$ -functions lying in the Chadrasekharan-Narasimhan class that may obey Riemann hypothesis.

Theorem 3.10.

Conjecture 1.3 holds true.

As we know that the Epstein zeta function $\zeta_{id}(2,s)$ and $L$ -functions $L(f,s)$ associated to Hecke eigenforms are particular examples of nice $L$ -functions that are lying in the Chadrasekharan-Narasimhan class. Thus, as an immediate implication, we have the following results.

Corollary 3.11.

Let $\zeta_{id}(2,s)$ be the Epstein zeta function defined in Example 4. For $k\geq 1,\ell>0$ , the Riemann hypothesis for $\zeta_{id}(2,s)$ is equivalent to

\displaystyle\sum_{n=1}^{\infty}\frac{r_{2}^{-1}(n)}{n^{k}}\exp\left(-\frac{x}% {n^{\ell}}\right)=O_{\epsilon,k,\ell}\left(x^{-\frac{k}{\ell}+\frac{1}{2\ell}+% \epsilon}\right),

for any $\epsilon>0$ , as $x\rightarrow\infty$ .

Corollary 3.12.

Let $f$ be a holomorphic Hecke eigenform of weight $\omega$ over the full modular group $SL_{2}(\mathbb{Z})$ and $L(f,s)$ be the $L$ -function defined in (2.6). For $k\geq\frac{\omega+1}{2},\ell>0$ , the Riemann hypothesis for $L(f,s)$ is equivalent to the following bound:

\displaystyle\sum_{n=1}^{\infty}\frac{\lambda_{f}^{-1}(n)}{n^{k}}\exp\left(-% \frac{x}{n^{\ell}}\right)=O_{\epsilon,k,f,\ell}\left(x^{-\frac{k}{\ell}+\frac{% 1}{2\ell}+\epsilon}\right),

for any $\epsilon>0$ as $x\rightarrow\infty$ .

In the next result, we state a new equivalent criteria for the Riemann hypothesis of $\zeta(s)$ . This criteria significantly differ from the previous known equivalent criteria for RH due to Hardy-Littlewood and Riesz. The heuristic of this bound comes from the identity mentioned in Theorem 3.5.

Theorem 3.13.

Let $r$ be a non-negative integer and $k\geq r+1,\ell>0$ . The Riemann hypothesis for $\zeta(s)$ is equivalent to the bound

\displaystyle P_{k,\sigma,\ell}(x):=\sum_{n=1}^{\infty}\frac{\sigma_{r}^{-1}(n% )}{n^{k}}\exp\left(-\frac{x}{n^{\ell}}\right)=O_{\epsilon,k,r,\ell}\left(x^{-% \frac{k}{\ell}+\frac{1+2r}{2\ell}+\epsilon}\right),

(3.5)

for any $\epsilon>0$ , as $x\rightarrow\infty$ .

4. Nuts and Bolts

In this section, we state a few well-known results, which will play a vital role in proving our main identity. We begin with an important asymptotic formula, namely, stirling’s formula which gives information about the asymptotic behaviour of $\Gamma(s)$ .

Lemma 4.1.

In a vertical strip $r_{1}\leq\sigma\leq r_{2}$ , we have

|\Gamma(\sigma+iT)|=\sqrt{2\pi}|T|^{\sigma-1/2}e^{-\frac{1}{2}\pi|T|}\left(1+O% \left(\frac{1}{|T|}\right)\right),\quad{\rm as}\quad|T|\rightarrow\infty.

Next, we document Euler’s summation formula, which will be useful in deriving the equivalent criteria for the Riemann hypothesis of the $L$ -functions.

Lemma 4.2.

Let $\{a_{n}\}$ be a sequence of complex numbers and $f(t)$ be a continuously differentiable function on $[1,x]$ . Consider $A(x):=\sum_{1\leq n\leq x}a_{n}$ . Then we have

\displaystyle\sum_{1\leq n\leq x}a_{n}f(n)=A(x)f(x)-\int_{1}^{x}A(t)f^{\prime}% (t){\rm d}t.

Proof.

A proof of this result can be found in [21, p. 17]. ∎

The next result gives an information about the behaviour of general $L$ -functions in a vertical strip.

Lemma 4.3.

Any $L$ -function is polynomially bounded in vertical strips $\sigma_{1}<\operatorname{Re}(s)<\sigma_{2},$ and $\operatorname{Im}(s)\geq 1$ .

Proof.

We refer [18, Lemma 5.2] for the proof. ∎

The forthcoming result will serve as one of the vital components in establishing equivalent criteria for the grand Riemann hypothesis.

Lemma 4.4.

Let $L(f,s)$ be a nice $L$ -function and its inverse defined in (1.10) and (1.11), respectively. For $k\geq 1$ , $\ell>0$ , we define the function

\displaystyle P_{k,f,\ell}(x):=\sum_{n=1}^{\infty}\frac{\mu_{f}(n)}{n^{k}}\exp% \left(-\frac{x}{n^{\ell}}\right).

Then for $\operatorname{Re}(s)>\frac{1-k}{\ell}$ , except for $s=0,1,2\cdots$ , we have

\displaystyle\int_{0}^{\infty}x^{-s-1}P_{k,f,\ell}(x)\mathrm{d}x=\frac{\Gamma(% -s)}{L(f,\ell s+k)}.

Proof.

We start with the definition of $P_{k,f,\ell}(x)$ . For $k\geq 1$ , we can write

$\displaystyle P_{k,f,\ell}(x)=\sum_{n=1}^{\infty}\frac{\mu_{f}(n)}{n^{k}}\exp% \left(-\frac{x}{n^{\ell}}\right)$	$\displaystyle=\sum_{n=1}^{\infty}\frac{\mu_{f}(n)}{n^{k}}\sum_{m=0}^{\infty}% \frac{(-1)^{m}x^{m}}{n^{\ell m}m!}$
	$\displaystyle=\sum_{m=0}^{\infty}\frac{(-1)^{m}x^{m}}{m!}\sum_{n=1}^{\infty}% \frac{\mu_{f}(n)}{n^{k+\ell m}}$
	$\displaystyle=\sum_{m=0}^{\infty}\frac{(-1)^{m}x^{m}}{m!L(f,\ell m+k)}$
	$\displaystyle=\frac{1}{L(f,k)}+\sum_{m=1}^{\infty}\frac{(-1)^{m}x^{m}}{m!L(f,% \ell m+k)}.$	(4.1)

For $\operatorname{Re}(s)>\frac{1-k}{\ell}$ , use the series expansion (1.10) of $L(f,k+\ell s)$ to write

$\displaystyle L(f,k+\ell s)\int_{0}^{\infty}x^{-s-1}P_{k,f,\ell}(x)\mathrm{d}x$	$\displaystyle=\sum_{n=1}^{\infty}\frac{A_{f}(n)}{n^{k}}\int_{0}^{\infty}\frac{% x^{-s-1}}{n^{\ell s}}P_{k,f,\ell}(x)\mathrm{d}x$
	$\displaystyle=\sum_{n=1}^{\infty}\frac{A_{f}(n)}{n^{k}}\int_{0}^{\infty}x^{-s-% 1}P_{k,f,\ell}\left(\frac{x}{n^{\ell}}\right)\mathrm{d}x$
	$\displaystyle=\int_{0}^{\infty}x^{-s-1}\sum_{n=1}^{\infty}\frac{A_{f}(n)}{n^{k% }}P_{k,f,\ell}\left(\frac{x}{n^{\ell}}\right)\mathrm{d}x.$	(4.2)

Here, in the second last step, we changed the variable $x$ by $\frac{x}{n^{\ell}}$ and interchange of the order of summation and integration in the last step is facilitated due to dominated convergence theorem. This is because the infinite sum is bounded by $e^{-x}$ , see (Proof.), which is integrable in the interval $(0,\infty)$ . Now, utilize (Proof.) to obtain

	$\displaystyle\sum_{n=1}^{\infty}\frac{A_{f}(n)}{n^{k}}P_{k,f,\ell}\left(\frac{% x}{n^{\ell}}\right)$	$\displaystyle=\sum_{n=1}^{\infty}\frac{A_{f}(n)}{n^{k}}\left(\frac{1}{L(f,k)}+% \sum_{m=1}^{\infty}\frac{(-1)^{m}x^{m}}{m!L(f,k+\ell m)n^{\ell m}}\right)$
		$\displaystyle=\begin{cases}e^{-x}-1,&\text{if}~{}k=1~{}\text{and}~{}L(f,s)~{}% \text{has a pole at}~{}s=1,\\ e^{-x},&\text{otherwise}.\end{cases}$		(4.3)

In this context, combining (Proof.) and (Proof.), when $k=1$ and $L(f,s)$ has a pole at $s=1$ , we obtain the following expression

	$\displaystyle L(f,k+\ell s)\int_{0}^{\infty}x^{-s-1}P_{k,f,\ell}(x)\mathrm{d}x$	$\displaystyle=\int_{0}^{\infty}x^{-s-1}(e^{-x}-1)\mathrm{d}x$
		$\displaystyle=\frac{-1}{s}\int_{0}^{\infty}x^{(1-s)-1}e^{-x}\mathrm{d}x\$
		$\displaystyle=-\frac{\Gamma(1-s)}{s}.$

This result is valid for $0<\operatorname{Re}(s)<1$ . Note that we used the fact that $\lim_{x\rightarrow 0}\frac{e^{-x}-1}{x^{s}}=0$ for $0<\operatorname{Re}(s)<1$ . Furthermore, we know $-\frac{\Gamma(1-s)}{s}=\Gamma(-s)$ . In other case, we can see that

\displaystyle L(f,k+\ell s)\int_{0}^{\infty}x^{-s-1}P_{k,f,\ell}(x)\mathrm{d}x% =\int_{0}^{\infty}x^{-s-1}e^{-x}\mathrm{d}x=\Gamma(-s),

valid for $\frac{1-k}{\ell}<\operatorname{Re}(s)<0$ . By analytic continuation, we can extend this to right half plane $\operatorname{Re}(s)>\frac{1-k}{\ell}$ except $s=0,1,2\cdots$ . The proof is now complete. ∎

Lemma 4.5.

Let $r$ be a non-negative integer and $k\geq r+1$ be any positive real number. We define

\displaystyle P_{k,\sigma,\ell}(x):=\sum_{n=1}^{\infty}\frac{\sigma_{r}^{-1}(n% )}{n^{k}}\exp\left(-\frac{x}{n^{\ell}}\right).

Then for $Re(s)>\frac{1-k+r}{\ell}$ except $s=0,1,2\cdots$ , we have

\displaystyle\int_{0}^{\infty}x^{-s-1}P_{k,\sigma,\ell}(x)\mathrm{d}x=\frac{% \Gamma(-s)}{\zeta(k+\ell s)\zeta(k+\ell s-r)}.

Proof.

The proof follows along the same line as of Lemma 4.4. Thus, we skip the details here. ∎

Now we state a result for the growth of the summatory function of an arithmetical function.

Lemma 4.6.

Let $a(n)$ be an arithmetical function such that $a(n)=O(\Phi(n))$ , where $\Phi(n)$ is an increasing function for $x\geq x_{0}$ . Suppose that the Dirichlet series $A(s)=\sum_{n=1}^{\infty}\frac{a(n)}{n^{s}}$ is absolutely convergent for $\operatorname{Re}(s)>c_{0}$ and

\displaystyle\sum_{n=1}^{\infty}\frac{a(n)}{n^{\sigma}}\leq\frac{1}{(\sigma-c_% {0})^{\gamma}}\quad\rm{as}\quad\sigma\rightarrow c_{0}^{+},

for some $\gamma>0$ . Then for any $c>c_{0}$ , one has

	$\displaystyle\sum_{n\leq x}a(n)$	$\displaystyle=\frac{1}{2\pi i}\int_{c-iT}^{c+iT}A(w)x^{w}\frac{\mathrm{d}w}{w}% +O\left(\frac{x^{c}}{T(c-c_{0})^{\gamma}}\right)$
		$\displaystyle+O\left(\frac{\Phi(2x)x\log(2x)}{T}\right)+O\left(\Phi(2x)\right),$

where $T$ is some large positive real number.

Proof.

This result can be found in [17, p. 486, Equation (A.10)]. ∎

Using the above identity, under the assumption of the Riemann hypothesis for $\zeta(s)$ , we obtain the below result which is an analogue of Littlewood’s bound (1.12).

Lemma 4.7.

Let $r$ be a non-negative integer. Under the assumption of the Riemann hypothesis for $\zeta(s)$ , we have

\displaystyle\sum_{n\leq x}\sigma_{r}^{-1}(n)=O\left(x^{\frac{1}{2}+r+\epsilon% }\right).

Proof.

First, assuming RH for the Riemann zeta function, we will prove this bound. In Lemma 4.6, we consider $a(n)=\sigma_{r}^{-1}(n)=\mu N^{r}*\mu$ . Then

\displaystyle|a(n)|=\bigg{|}\sum_{d|n}\mu(d)d^{r}\mu\left(\frac{n}{d}\right)% \bigg{|}\leq\sum_{d|n}d^{r}\leq\sum_{d|n}n^{r}=n^{r}d(n)=O(n^{r+\epsilon}),

for any $\epsilon>0$ . This suggests us to take $\Phi(n)=n^{r+\epsilon}$ . Thus, with $A(s)=\frac{1}{\zeta(s)\zeta(s-r)}$ and $c=1+r+\epsilon$ , we see that

	$\displaystyle\sum_{n\leq x}\sigma_{r}^{-1}(n)$	$\displaystyle=\frac{1}{2\pi i}\int_{1+r+\epsilon-iT}^{1+r+\epsilon+iT}\frac{1}% {\zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w+O\left(\frac{x^{1+r+\epsilon}}{% T}\right)$
		$\displaystyle+O\left(\frac{x^{r+\epsilon}x\log(2x)}{T}\right)+O\big{(}x^{r+% \epsilon}\big{)}.$

Now we change the line of integration from $\operatorname{Re}(s)=1+r+\epsilon$ to $\operatorname{Re}(s)=\frac{1}{2}+r+\epsilon$ . Then we have

	$\displaystyle\sum_{n\leq x}\sigma_{r}^{-1}(n)$	$\displaystyle=\frac{1}{2\pi i}\bigg{[}\int_{\frac{1}{2}+r+\epsilon+iT}^{1+r+% \epsilon+iT}+\int_{\frac{1}{2}+r+\epsilon-iT}^{\frac{1}{2}+r+\epsilon+iT}+\int% _{1+r+\epsilon-iT}^{\frac{1}{2}+r+\epsilon-iT}\bigg{]}\frac{1}{\zeta(w)\zeta(w% -r)}\frac{x^{w}}{w}\mathrm{d}w$
		$\displaystyle+O\left(\frac{x^{1+r+\epsilon}}{T}\right)+O(x^{r+\epsilon}).$		(4.4)

Under the assumption of RH, we know $\frac{1}{\zeta(\sigma+it)}=O(|t|^{\epsilon})$ for $\sigma>1/2$ . Using this bound and under simplification, one can check that

	$\displaystyle\left\|\int_{\frac{1}{2}+r+\epsilon+iT}^{1+r+\epsilon+iT}\frac{1}{% \zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w\right\|$	$\displaystyle\ll\int_{\frac{1}{2}+r+\epsilon}^{1+r+\epsilon}x^{\sigma}T^{2% \epsilon-1}d\sigma\ll T^{2\epsilon-1}x^{2+r+\epsilon},$
	$\displaystyle\left\|\int_{\frac{1}{2}+r+\epsilon-iT}^{\frac{1}{2}+r+\epsilon+iT% }\frac{1}{\zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w\right\|$	$\displaystyle\ll\int_{-T}^{T}\|t\|^{2\epsilon}x^{\frac{1}{2}+r+\epsilon}\frac{dt% }{\|t\|}\ll T^{2\epsilon}x^{\frac{1}{2}+r+\epsilon},$
	$\displaystyle\left\|\int_{1+r+\epsilon-iT}^{\frac{1}{2}+r+\epsilon-iT}\frac{1}{% \zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w\right\|$	$\displaystyle\ll\int_{1+r+\epsilon}^{\frac{1}{2}+r+\epsilon}x^{\sigma}T^{2% \epsilon-1}d\sigma\ll T^{2\epsilon-1}x^{2+r+\epsilon}.$

Utilizing these bounds in (Proof.), we arrive

	$\displaystyle\sum_{n\leq x}\sigma_{r}^{-1}(n)$	$\displaystyle=O\left(T^{2\epsilon-1}x^{2+r+\epsilon}\right)+O\left(T^{2% \epsilon}x^{\frac{1}{2}+r+\epsilon}\right)+O\left(\frac{x^{1+r+\epsilon}}{T}% \right)+O\big{(}x^{r+\epsilon}\big{)}$
		$\displaystyle=O\left(T^{2\epsilon-1}x^{2+r+\epsilon}\right)+O\left(T^{2% \epsilon}x^{\frac{1}{2}+r+\epsilon}\right).$

Taking $T=x^{\frac{3}{2}}$ , the result follows.

∎

Now we define an important special function, namely, the Meijer $G$ -function [22, p. 415, Definition 16.17], which is a generalization of many well-known special functions. Let $m,n,p,q$ be non-negative integers with $0\leq m\leq q$ , $0\leq n\leq p$ . Let $a_{1},\cdots,a_{p}$ and $b_{1},\cdots,b_{q}$ be complex numbers with $a_{i}-b_{j}\not\in\mathbb{N}$ for $1\leq i\leq n$ and $1\leq j\leq m$ . Then the Meijer $G$ -function is defined by

\displaystyle G_{p,q}^{\,m,n}\!\left(\,\begin{matrix}a_{1},\cdots,a_{p}\\ b_{1},\cdots,b_{q}\end{matrix}\;\Big{|}z\right):=\frac{1}{2\pi i}\int_{L}\frac% {\prod_{j=1}^{m}\Gamma(b_{j}-s)\prod_{j=1}^{n}\Gamma(1-a_{j}+s)z^{s}}{\prod_{j% =m+1}^{q}\Gamma(1-b_{j}+s)\prod_{j=n+1}^{p}\Gamma(a_{j}-s)}\mathrm{d}s,

(4.5)

where the line of integration $L$ , from $-i\infty$ to $+i\infty$ , separates the poles of the factors $\Gamma(1-a_{j}+s)$ from those of the factors $\Gamma(b_{j}-s)$ . The above integral converges if $p+q<2(m+n)$ and $|\arg(z)|<\left(m+n-\frac{p+q}{2}\right)\pi$ .

Next, we state Slater’s theorem [22, p. 415, Equation 16.17.2], which will allow us to express Meijer $G$ -function in terms of generalized hypergeometric functions. If $p\leq q$ and $b_{j}-b_{k}\not\in\mathbb{Z}$ for $j\neq k$ , $1\leq j,k\leq m$ , then

		$\displaystyle G_{p,q}^{\,m,n}\!\left(\,\begin{matrix}a_{1},\cdots,a_{p}\\ b_{1},\cdots,b_{q}\end{matrix}\;\Big{\|}z\right)$
		$\displaystyle\quad=\sum_{k=1}^{m}A_{p,q,k}^{m,n}(z){}_{p}F_{q-1}\left(\begin{% matrix}1+b_{k}-a_{1},\cdots,1+b_{k}-a_{p}\\ 1+b_{k}-b_{1},\cdots,*,\cdots,1+b_{k}-b_{q}\end{matrix}\Big{\|}(-1)^{p-m-n}z% \right),$		(4.6)

where $*$ means that the entry $1+b_{k}-b_{k}$ is omitted and

\displaystyle A_{p,q,k}^{m,n}(z):=\frac{z^{b_{k}}\prod_{j=1,j\neq k}^{m}\Gamma% (b_{j}-b_{k})\prod_{j=1}^{n}\Gamma(1+b_{k}-a_{j})}{\prod_{j=m+1}^{q}\Gamma(1+b% _{k}-b_{j})\prod_{j=n+1}^{p}\Gamma(a_{j}-b_{k})}.

5. Proof of Main Result

Proof of Theorem 3.1.

Shifting the line of integration from $\operatorname{Re}(s)>0$ to $\operatorname{Re}(s)=d\in(-1,0)$ in the inverse mellin transform of $\Gamma(s)$ , we have

\displaystyle e^{-x}-1=\frac{1}{2\pi i}\int_{d-i\infty}^{d+i\infty}\Gamma(s)x^% {-s}{\rm d}s.

(5.1)

We choose $k_{0}$ such that the following series

\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k}},\quad\sum_{n=1}^{\infty}\frac{b^{-1% }(n)}{n^{k}}

are convergent for $k=k_{0}$ and absolutely convergent for all $k>k_{0}$ . Then for $k\geq k_{0}$ , $x>0$ , we use (5.1) to write

$\displaystyle\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k}}\exp\left({-\frac{x}{n^% {1/A}}}\right)$	$\displaystyle=\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k}}+\sum_{n=1}^{\infty}% \frac{a^{-1}(n)}{n^{k}}\left(\exp\left({-\frac{x}{n^{1/A}}}\right)-1\right)$
	$\displaystyle=\frac{1}{\phi(k)}+\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k}}% \frac{1}{2\pi i}\int_{d-i\infty}^{d+i\infty}\Gamma(s)\left(\frac{x}{n^{1/A}}% \right)^{-s}{\rm d}s$
	$\displaystyle=\frac{1}{\phi(k)}+\frac{1}{2\pi i}\int_{d-i\infty}^{d+i\infty}% \frac{\Gamma(s)}{\phi(k-s/A)}x^{-s}{\rm d}s.$	(5.2)

Since $\operatorname{Re}(k-s/A)>k_{0}$ , hence $1/\phi(k-s/A)=\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k-s/A}}$ is absolutely and uniformly convergent in any compact subset of the domain. Thus, interchanging the order of summation and integration is justified. We now intend to solve the following vertical integral,

\displaystyle J_{k,A}(x):=\frac{1}{2\pi i}\int_{d-i\infty}^{d+i\infty}\frac{% \Gamma(s)}{\phi(k-s/A)}x^{-s}{\rm d}s.

We shall investigate the poles of the integrand function in the first place. One can observe that $\Gamma(s)$ has simple poles at non-positive integers. Let $S$ be the set of all the trivial zeros of $\phi(s)$ and to collect the contribution of the residual term due to non-trivial zeros of $\phi(k-s/A)$ , we change the line of integration from $\operatorname{Re}(s)=d\in(-1,0)$ to $\operatorname{Re}(s)=\lambda\in(Ak+B,Ak+B+\epsilon)$ with $0<\epsilon<1$ . For that, we consider a rectangular contour $\mathcal{C}$ determined by the line segments $[\lambda-iT,\lambda+iT],[\lambda+iT,d+iT],[d+iT,d-iT]$ , and $[d-iT,\lambda-iT]$ , where $T$ is some large positive number.

We now use Cauchy’s residue theorem to get

\displaystyle\frac{1}{2\pi i}\int_{\mathcal{C}}\frac{\Gamma(s)}{\phi(k-s/A)}x^% {-s}{\rm d}s=R_{0}+R_{t}+\sum_{|\operatorname{Im}(\rho)|<T}{R}_{\rho}(x),

(5.3)

where $R_{0}$ stands for the residual term corresponding to the simple pole of $\Gamma(s)$ at $s=0$ , $R_{t}$ is the residual term due to trivial zeros of $\phi(k-s/A)$ with $0\leq\operatorname{Re}\left(k-\frac{s}{A}\right)\leq Ak+B$ and ${R}_{\rho}(x)$ is the residual term corresponding to the non-trivial zero $\rho$ of $\phi(k-s/A)$ . Next goal is to show that the horizontal integrals

	$\displaystyle H_{1}(T,A)=\frac{1}{2\pi i}\int_{\lambda+iT}^{d+iT}\frac{\Gamma(% s)}{\phi(k-s/A)}x^{-s}{\rm d}s,$
	$\displaystyle H_{2}(T,A)=\frac{1}{2\pi i}\int_{d-iT}^{\lambda-iT}\frac{\Gamma(% s)}{\phi(k-s/A)}x^{-s}{\rm d}s,$

will vanish as $T\rightarrow\infty$ . This follows if we make use of Stirling’s formula for $\Gamma(s)$ , that is, Lemma 4.1 along with Lemma 4.3. Thus, letting $T\rightarrow\infty$ in (5.3), we arrive at

\displaystyle\frac{1}{2\pi i}\int_{d-i\infty}^{d+i\infty}\frac{\Gamma(s)}{\phi% (k-s/A)}x^{-s}{\rm d}s=\frac{1}{2\pi i}\int_{\lambda-i\infty}^{\lambda+i\infty% }\frac{\Gamma(s)}{\phi(k-s/A)}x^{-s}{\rm d}s-R_{0}-R_{t}-\sum_{\rho}{R}_{\rho}% (x),

(5.4)

where the sum over $\rho$ runs through all the non-trivial zeros of $\phi(k-s/A)$ . The residue at $s=0$ can be evaluated as

\displaystyle R_{0}=\lim_{s\rightarrow 0}\frac{s\Gamma(s)}{\phi(k-s/A)}x^{-s}=% \frac{1}{\phi(k)}.

(5.5)

Considering the non-trivial zeros of $\phi(s)$ to be simple, we get

\displaystyle\sum_{\rho}{R}_{\rho}(x)=\sum_{\rho}\lim_{s\rightarrow A(k-\rho)}% \frac{\left(s-A(k-\rho)\right)\Gamma(s)}{\phi(k-s/A)}x^{-s}=-A\sum_{\rho}\frac% {\Gamma(A(k-\rho))}{\phi^{\prime}(\rho)}x^{-A(k-\rho)}.

(5.6)

Substituting (5.4), (5.5) and (5.6) in ( Proof of Theorem 3.1.), we get

	$\displaystyle\sum_{n=1}^{\infty}\frac{a^{-1}(n)}{n^{k}}\exp\left({-\frac{x}{n^% {1/A}}}\right)$	$\displaystyle=\frac{1}{2\pi i}\int_{\lambda-i\infty}^{\lambda+i\infty}\frac{% \Gamma(s)}{\phi(k-s/A)}x^{-s}{\rm d}s-R_{t}$
		$\displaystyle+A\sum_{\rho}\frac{\Gamma(A(k-\rho))}{\phi^{\prime}(\rho)}x^{-A(k% -\rho)}.$		(5.7)

We shall now focus on the derivation of the line integral present above. Let us define

\displaystyle I_{k,A}(x):=\frac{1}{2\pi i}\int_{\lambda-i\infty}^{\lambda+i% \infty}\frac{\Gamma(s)}{\phi\left(k-\frac{s}{A}\right)}x^{-s}{\rm d}s.

(5.8)

Here, we bring into play the role of the functional equation for $\phi(s)$ . Replacing $s$ by $k-s/A$ in (2.4), it can be seen that

\displaystyle\frac{1}{\phi(k-\frac{s}{A})}=\frac{c^{2k-\delta-\frac{2s}{A}}% \Gamma(Ak+B-s)}{\nu\Gamma(A\delta-Ak+B+s)\psi(\delta-k+\frac{s}{A})}.

(5.9)

Utilizing (5.9) in (5.8), we get

\displaystyle I_{k,A}(x)=\frac{1}{\nu}\frac{c^{2k-\delta}}{2\pi i}\int_{% \lambda-i\infty}^{\lambda+i\infty}\frac{\Gamma(s)\Gamma(Ak+B-s)}{\Gamma(A(% \delta-k)+B+s)\psi(\delta-k+\frac{s}{A})}(c^{2/A}x)^{-s}{\rm d}s.

(5.10)

Note that $\operatorname{Re}(\delta-k+\frac{s}{A})>\delta+\frac{B}{A}$ . Thus, to use the series expansion of $\frac{1}{\psi(\delta-k+\frac{s}{A})}=\sum_{n=1}^{\infty}\frac{b^{-1}(n)}{n^{s/% A}}n^{k-\delta}$ in (5.10), constraints us to choose $k_{0}$ such that $k_{0}=\delta+\frac{B}{A}$ . Then interchanging the order of summation and integration leads us to obtain that,

\displaystyle I_{k,A}(x)=\frac{c^{2k-\delta}}{\nu}\sum_{n=1}^{\infty}\frac{b^{% -1}(n)}{n^{\delta-k}}V_{k,A}(X_{n}),

(5.11)

where

\displaystyle V_{k,A}(X_{n})=\frac{1}{2\pi i}\int_{\lambda-i\infty}^{\lambda+i% \infty}\frac{\Gamma(s)\Gamma(Ak+B-s)}{\Gamma(A(\delta-k)+B+s)}X_{n}^{-s}{\rm d% }s,

(5.12)

and $X_{n}=(c^{2}n)^{1/A}x$ . We shall now focus on $V_{k,A}(X_{n})$ and we aim at writing this vertical integral in terms of the Meijer $G$ -function. However, it can be seen that the line of integration $Ak+B<\operatorname{Re}(s)=\lambda<Ak+B+\epsilon$ does not separate the poles of $\Gamma(s)$ from that of $\Gamma(Ak+B-s)$ . Thus, to use the definition of the Meijer $G$ -function (4.5), we relocate the line of integration from $\operatorname{Re}(s)=\lambda$ to $0<\operatorname{Re}(s)=\beta<Ak+B$ to ensure that the line of integration separates the poles of $\Gamma(s)$ and $\Gamma(Ak+B-s)$ . Finally, we consider a contour $\mathcal{D}_{1}$ determined by the line segments $[\lambda-iT,\lambda+iT],[\lambda+iT,\beta+iT],[\beta+iT,\beta-iT]$ , and $[\beta-iT,\lambda-iT]$ .

Again, utilizing the Cauchy’s residue theorem, we have

\displaystyle\frac{1}{2\pi i}\left(\int_{\lambda-iT}^{\lambda+iT}+\int_{% \lambda+iT}^{\beta+iT}+\int_{\beta+iT}^{\beta-iT}+\int_{\beta-iT}^{\lambda-iT}% \right)\frac{\Gamma(s)\Gamma(Ak+B-s)}{\Gamma(A(\delta-k)+B+s)}X_{n}^{-s}{\rm d% }s=R_{Ak+B},

(5.13)

where $R_{Ak+B}$ denotes the residual term due to the simple pole of $\Gamma(Ak+B-s)$ at $s=Ak+B$ . We can easily verify that

\displaystyle R_{Ak+B}=-\frac{\Gamma(Ak+B)}{\Gamma(A\delta+2B)}X_{n}^{-Ak-B}.

(5.14)

Letting $T\rightarrow\infty$ , one can show that horizontal integrals vanish. Thus substituting (5.14) in (5.13) and using (5.12), it leads us to obtain that

\displaystyle V_{k,A}(X_{n})=\frac{1}{2\pi i}\int_{\beta-i\infty}^{\beta+i% \infty}\frac{\Gamma(s)\Gamma(Ak+B-s)}{\Gamma(A(\delta-k)+B+s)}X_{n}^{-s}{\rm d% }s-\frac{\Gamma(Ak+B)}{\Gamma(A\delta+2B)}X_{n}^{-Ak-B}.

(5.15)

As we know that the line of integration $\operatorname{Re}(s)=\beta$ separates the poles of $\Gamma(s)$ from the poles of $\Gamma(Ak+B-s)$ , so we are all set to write the vertical integral, in (5.15), in terms of the Meijer $G$ -function (4.5) with $n=p=m=1,q=2$ and $a_{1}=1,b_{1}=Ak+B$ and $b_{2}=1-A\delta+Ak-B$ . Thus, we have

\displaystyle\frac{1}{2\pi i}\int_{\beta-i\infty}^{\beta+i\infty}\frac{\Gamma(% s)\Gamma(Ak+B-s)}{\Gamma(A(\delta-k)+B+s)}X_{n}^{-s}{\rm d}s=G_{1,2}^{1,1}% \left(\begin{matrix}1\\ Ak+B,1-A\delta+Ak-B\end{matrix}\Big{|}\frac{1}{X_{n}}\right).

(5.16)

Here one can easily check that $p+q<2(m+n)$ and $|\arg(\frac{1}{X_{n}})|<\left(m+n-\frac{p+q}{2}\right)\pi$ . Thus all the necessary conditions for the convergence of the above Meijer $G$ -function are satisfied. Now invoking Slater’s theorem (4), one can see that

\displaystyle G_{1,2}^{1,1}\left(\begin{matrix}1\\ Ak+B,1-A\delta+Ak-B\end{matrix}\Big{|}\frac{1}{X_{n}}\right)=\frac{1}{X_{n}^{% Ak+B}}\frac{\Gamma(Ak+B)}{\Gamma(A\delta+2B)}{}_{1}F_{1}\left(\begin{matrix}Ak% +B\\ A\delta+2B\end{matrix}\Big{|}-\frac{1}{X_{n}}\right).

(5.17)

Finally, substituting (5.17) in (5.16) and together with (5.15), (5.11), (5.8) and then combining all these terms in (5), we complete the proof of Theorem 3.1. ∎

Proof of Corollary 3.2.

Letting $k=\delta+\frac{B}{A}$ and $x=\beta>0$ with $\alpha\beta=\frac{1}{c^{2/A}}$ in Theorem 3.1 and then multiplying by $\sqrt{\nu}\beta^{\frac{A\delta+2B}{B}}$ throughout and using the fact that $|\nu|=1$ , one can complete the proof of this result. ∎

Proof of Corollary 3.3.

The identity (3.2) represents a particular instance of the identity (3.1) with $a(n)=b(n)=r_{2}(n)$ , where $r_{2}(n)$ is defined in Example 4 for $j=2$ . Note that $\zeta_{id}(2,s)$ satisfies the following functional equation

\displaystyle\pi^{-s}\Gamma(s)\zeta_{id}(2,s)=\pi^{-(1-s)}\Gamma\left(1-s% \right)\zeta_{id}\left(2,1-s\right).

This corresponds to (2.4) for $a(n)=b(n)=r_{2}(n),c=\frac{1}{\pi},A=1,B=0,\delta=1$ and $\nu=1.$ Now substituting these values in Theorem 3.1 and noting that the term $R_{t}$ is zero as we do not encounter any trivial zeros of $\zeta_{id}(2,k-s)$ with $0\leq\operatorname{Re}(k-s)\leq k$ , and also $\sum_{n=1}^{\infty}\frac{r_{2}^{-1}(n)}{n}=0$ since we know the relation (2.8), we complete the proof. ∎

Proof of Corollary 3.4.

This scenario represents a special case of (3.1) for the Dedekind zeta function for the imaginary quadratic field $Q(\sqrt{-D})$ , which is defined in Example 5. From the functional equation (2.10), it is clear that $c=\frac{\sqrt{D}}{2\pi},A=1,B=0,\delta=1$ and $\nu=1$ . Now putting these values in Theorem 3.1 and observing that $R_{t}=0$ since we do not encounter a trivial zero of $\zeta_{Q(\sqrt{-D})}(k-s)$ with $0\leq\operatorname{Re}(k-s)\leq k$ , and $\sum_{n=1}^{\infty}\frac{\mathsf{a}_{n}^{-1}}{n}=0$ , one can complete the proof. ∎

Proof of Theorem 3.5.

This identity is a special case of the identity (3.1) with $a(n)=b(n)=\sigma_{r}(n),c=\frac{1}{2\pi},A=1,B=0,\delta=r+1$ and $\nu=(-1)^{\frac{r+1}{2}}$ follows from Example 1. In this case $\phi(s)=\zeta(s)\zeta(s-r)$ , where $r\geq 1$ odd integer. Thus, the trivial zeros of $\phi(s)$ are at $\{-2m,r-2m\}$ , with $m\in\mathbb{N}$ . Hence, the trivial zeros of $\phi(k-s)$ with $0\leq\operatorname{Re}(k-s)\leq k$ are at $k-r+2m$ such that $1\leq m\leq\lfloor\frac{r}{2}\rfloor$ . Therefore, from (5.3), one can see that the residual term $R_{t}$ becomes

$\displaystyle R_{t}$	$\displaystyle=\sum_{m=1}^{\lfloor\frac{r}{2}\rfloor}\lim_{s\rightarrow k-r+2m}% \frac{(s-(k-r+2m))\Gamma(s)x^{-s}}{\zeta(k-s)\zeta(k-s-r)}$
	$\displaystyle=-\sum_{m=1}^{\lfloor\frac{r}{2}\rfloor}\frac{\Gamma(k-r+2m)x^{r-% k-2m}}{\zeta(r-2m)\zeta^{{}^{\prime}}(-2m)}$
	$\displaystyle=\sum_{m=1}^{[\frac{r}{2}]}\frac{\Gamma(k-r+2m)}{\zeta(r-2m)\zeta% (2m+1)}\frac{(-1)^{m}2(2\pi)^{2m}}{(2m)!}x^{r-k-2m}.$	(5.18)

In the final step we utilized the identity

\displaystyle\zeta^{{}^{\prime}}(-2m)=\frac{(-1)^{m}(2m)!\zeta(2m+1)}{2(2\pi)^% {2m}}.

From (5.3), one can see that the integrand function $\frac{\Gamma(s)x^{-s}}{\phi(k-s/A)}$ , and in this case $A=1$ and $\phi(k-s)=\zeta(k-s)\zeta(k-s-r)$ , so the residual term $R_{\rho}(x)$ will depend on the non-trivial zeros of $\zeta(k-s)$ as well as $\zeta(k-s-r)$ . Thus, if we assume that the non-trivial zeros of $\zeta(s)$ are simple, then one can show that

\displaystyle\sum_{\rho}R_{\rho}(x)=-\sum_{\rho}\frac{x^{\rho-k}}{\zeta^{{}^{% \prime}}(\rho)}\bigg{[}\frac{\Gamma(k-\rho)}{\zeta(\rho-r)}+\frac{x^{r}\Gamma(% k-r-\rho)}{\zeta(r+\rho)}\bigg{]},

(5.19)

where the sum over $\rho$ runs through the non-trivial zeros of $\zeta(s)$ . Again note that $\phi(s)=\zeta(s)\zeta(s-r)$ has a simple pole at $1+r$ , thus we will have $\sum_{n=1}^{\infty}\frac{\sigma^{-1}_{r}(n)}{n^{r+1}}=0$ . Substituting the above values (5.18)-(5.19) of $R_{t}$ and $R_{\rho}(x)$ in Theorem 3.1, we complete the proof of Theorem 3.5. ∎

Proof of Corollary 3.6.

Substituting $k=r+1$ , $x=\alpha$ , $\alpha\beta=4\pi^{2}$ and multiplying by $\alpha^{\frac{r+1}{2}}$ on both sides of Theorem 3.5, we obtain the result. ∎

Proof of Theorem 3.7.

This identity represents a particular instance of (3.1), for the normalized Fourier coefficients of the holomorphic Hecke eigenform $f(z)$ . Here, we take $a(n)=b(n)=\lambda_{f}(n)$ and the corresponding $L$ -function $L(f,s)$ is defined in Example 3. In this case, we encounter a pole of the integrand $\frac{\Gamma(s)x^{-s}}{L(f,k-s)}$ due to the trivial zero of $L(f,k-s)$ at $s=k+\frac{\omega-1}{2}$ which satisfies the condition $0\leq\operatorname{Re}(k-s)\leq k+\frac{w-1}{2}$ . Thus, we get our residual term $R_{t}$ to be

	$\displaystyle R_{t}$	$\displaystyle=\lim_{s\rightarrow k+\frac{\omega-1}{2}}\frac{\left(s-\left(k+% \frac{\omega-1}{2}\right)\right)\Gamma(s)x^{-s}}{L(f,k-s)}$
		$\displaystyle=\lim_{s\rightarrow k+\frac{\omega-1}{2}}\frac{\left(s-\left(k+% \frac{\omega-1}{2}\right)\right)\Gamma\left(k-s+\frac{\omega-1}{2}\right)% \Gamma(s)(2\pi)^{1+2s-2k}x^{-s}}{\Gamma\left(1-k+s+\frac{\omega-1}{2}\right)L(% f,1-k+s)}$
		$\displaystyle=-\frac{(2\pi)^{\omega}}{x^{k+\frac{\omega-1}{2}}}\frac{\Gamma% \left(k+\frac{\omega-1}{2}\right)}{\Gamma(\omega)}\frac{1}{L\left(f,\frac{% \omega+1}{2}\right)}.$

Here, in the penultimate step, we have used functional equation (2.7) for $L(f,s)$ . Substituting this value of $R_{t}$ along with $c=\frac{1}{2\pi},A=1,B=\frac{\omega-1}{2}$ , $\delta=1$ and $\nu=1$ in Theorem 3.1 and using $\frac{1}{L\left(f,\frac{\omega+1}{2}\right)}=\sum_{n=1}^{\infty}\frac{\lambda_% {f}^{-1}(n)}{n^{\frac{\omega+1}{2}}}$ , we obtain (3.3). ∎

Proof of Corollary 3.8.

Letting $k=\frac{\omega+1}{2}$ , $x=\beta$ , $\alpha\beta=4\pi^{2}$ in Theorem 3.7 and multiplying throughout by $\beta^{\frac{\omega}{2}}$ , we obtain the result. ∎

Proof of Corollary 3.9.

This result is an immediate implication of Corollary 3.8 for the Ramanujan cusp form of weight $12$ . Thus, substituting $\lambda_{f}(n)=\tau_{0}(n)$ and $\omega=12$ , we get our modular relation (3.4). ∎

Proof of Theorem 3.10.

Let $L(f,s)$ be a nice $L$ -function defined as in (1.10). First, we begin by presupposing that the grand Riemann hypothesis (GRH) for $L(f,s)$ holds true. Then from (1.12), for any $\epsilon>0$ , we have

\displaystyle G(x):=\sum_{1\leq n\leq x}\mu_{f}(n)\ll_{\epsilon,f}x^{\frac{1}{% 2}+\epsilon}.

(5.20)

Now, we employ Euler’s partial summation formula, i.e., Lemma 4.2 with $a(n)=\mu_{f}(n)$ and $f(t)=t^{-k}$ to see that

\displaystyle H(m,n):=\sum_{i=m}^{n}\frac{\mu_{f}(i)}{i^{k}}

\displaystyle=G(n)f(n)-G(m-1)f(m-1)-\int_{m-1}^{n}G(t)f^{\prime}(t){\rm d}t.

(5.21)

Utilizing (5.20) in (5.21), one can obtain

\displaystyle H(m,n)=O_{\epsilon,k}\left(m^{\frac{1}{2}-k+\epsilon}\right).

(5.22)

The above bound is uniform in $n$ . Our main objective is to determine the bound for the following infinite series, under the assumption of the GRH for $L(f,s)$ ,

\displaystyle P_{k,f,\ell}(x):=\sum_{n=1}^{\infty}\frac{\mu_{f}(n)}{n^{k}}\exp% \left(-\frac{x}{n^{\ell}}\right).

For simplicity, we replace $x$ by $x^{\ell}$ and separate the sum into finite and infinite part as

	$\displaystyle P_{k,f,\ell}(x^{\ell})$	$\displaystyle=\sum_{n=1}^{m-1}\frac{\mu_{f}(n)}{n^{k}}\exp\left(-\frac{x^{\ell% }}{n^{\ell}}\right)+\sum_{n=m}^{\infty}\frac{\mu_{f}(n)}{n^{k}}\exp\left(-% \frac{x^{\ell}}{n^{\ell}}\right)$
		$\displaystyle:=Q_{1}(x^{\ell})+Q_{2}(x^{\ell}).$		(5.23)

where $m=[x^{1-\epsilon}]+1$ . To get an estimate for $Q_{1}(x^{\ell})$ , we write

	$\displaystyle\|Q_{1}(x^{\ell})\|=\bigg{\|}\sum_{n=1}^{m-1}\frac{\mu_{f}(n)}{n^{k}% }\exp\left(-\frac{x^{\ell}}{n^{\ell}}\right)\bigg{\|}$	$\displaystyle\leq\sum_{n=1}^{m-1}\|H(n,n)\|\exp\left(-\frac{x^{\ell}}{m^{\ell}}\right)$
		$\displaystyle\ll\sum_{n=1}^{m-1}n^{\frac{1}{2}-k+\epsilon}\exp\left(-\frac{x^{% \ell}}{m^{\ell}}\right)$
		$\displaystyle\ll_{k,\epsilon}m^{\frac{3}{2}-k+\epsilon}\exp\left(-\frac{x^{% \ell}}{m^{\ell}}\right).$

This gives

\displaystyle Q_{1}(x^{\ell})=O\left(\exp(-x^{\ell\epsilon})x^{(1-\epsilon)(% \frac{3}{2}-k+\epsilon)}\right),

(5.24)

as $m-1=[x^{1-\epsilon}]$ . We shall now try to estimate $Q_{2}(x^{\ell})$ . For sufficiently large integer $N>m$ , we write

	$\displaystyle\sum_{n=m}^{N}\frac{\mu_{f}(n)}{n^{k}}\exp\left(-\frac{x^{\ell}}{% n^{\ell}}\right)$	$\displaystyle=\sum_{n=m}^{N-1}H(m,n)\left(\exp\left(-\frac{x^{\ell}}{n^{\ell}}% \right)-\exp\left(-\frac{x^{\ell}}{(n+1)^{\ell}}\right)\right)$
		$\displaystyle+H(m,n)\exp\left(-\frac{x^{\ell}}{N^{\ell}}\right).$

Allowing $N\rightarrow\infty$ and invoking the bound (5.22) for $H(m,n)$ , it can be inferred that

\displaystyle Q_{2}(x^{\ell})=Q_{3}(x^{\ell})+O_{\epsilon,k}\left(m^{\frac{1}{% 2}-k+\epsilon}\right),

(5.25)

where

\displaystyle Q_{3}(x^{\ell})=\sum_{n=m}^{N-1}H(m,n)\left(\exp\left(-\frac{x^{% \ell}}{n^{\ell}}\right)-\exp\left(-\frac{x^{\ell}}{(n+1)^{\ell}}\right)\right).

Now to simplify further, we use Cauchy’s mean value theorem with the function $T(z)=\exp\left(-\frac{x^{\ell}}{z^{\ell}}\right).$ One can find $z_{n}\in(n,n+1)$ such that

T(n)-T(n+1)=-T^{{}^{\prime}}(z_{n})=-\frac{\ell x^{\ell}}{z_{n}^{\ell+1}}\exp% \left(-\frac{x^{\ell}}{z_{n}^{\ell}}\right).

Employing the above fact in conjunction with (5.22), we obtain

$\displaystyle\|Q_{3}(x^{\ell})\|$	$\displaystyle\ll_{\epsilon,\ell}m^{\frac{1}{2}-k+\epsilon}\sum_{n=m}^{\infty}% \frac{x^{\ell}}{z_{n}^{\ell+1}}\exp\left(-\frac{x^{\ell}}{z_{n}^{\ell}}\right)$
	$\displaystyle\ll_{\epsilon,\ell}m^{\frac{1}{2}-k+\epsilon}\sum_{n=m}^{\infty}% \frac{x^{\ell}}{n^{\ell+1}}$
	$\displaystyle\ll_{\epsilon,\ell}x^{\frac{1}{2}-k+\epsilon}\frac{x^{\ell}}{m^{% \ell}}\ll_{\epsilon,\ell}x^{\frac{1}{2}-k+\epsilon^{\prime}}$	(5.26)

as $m\sim x^{1-\epsilon}$ . Now utilizing (5) in (5.25), we see that

\displaystyle|Q_{2}(x^{\ell})|=O_{\epsilon,k,\ell}\left(m^{\frac{1}{2}-k+% \epsilon}\right).

(5.27)

Finally, substituting the bounds (5.24) and (5.27) in (5), one can see that due to exponential decay, the bound for $Q_{1}(x^{\ell})$ goes to zero much faster than the bound for $Q_{2}(x^{\ell})$ as $x\rightarrow\infty$ , and therefore

\displaystyle P_{k,f,\ell}(x^{\ell})=O_{\epsilon,k,\ell}\left(x^{\frac{1}{2}-k% +\epsilon}\right).

Replace $x$ by $x^{1/\ell}$ , we get the desired bound (1.13).

Now, turning to the converse aspect, that is, assuming the validity of the bound (1.13), we aim to show that the GRH for $L(f,s)$ follows. Utilizing Lemma 4.4, we deduced that

\displaystyle L(f,\ell s+k)\int_{0}^{\infty}x^{-s-1}P_{k,f,\ell}(x)\mathrm{d}x% =\Gamma(-s),

(5.28)

valid for $\operatorname{Re}(s)>\frac{1-k}{\ell}$ except for non-negative integers. We aim to extend the validity of (5.28) towards the left half plane. To accomplish this, we choose a sufficiently large positive real number $R$ and rewrite the expression (5.28) as

\displaystyle L(f,\ell s+k)\left(\int_{0}^{R}+\int_{R}^{\infty}\right)x^{-s-1}% P_{k,f,\ell}(x)\mathrm{d}x=\Gamma(-s).

Utilizing the bound (1.13) for $P_{k,f,\ell}(x)$ , it is evident that unbounded part possesses analyticity for $\operatorname{Re}(s)>\frac{1}{2\ell}-\frac{k}{\ell}$ . Consequently, identity (5.28) is analytic in the strip $\frac{1}{2\ell}-\frac{k}{\ell}<\operatorname{Re}(s)<\frac{1-k}{\ell}$ . Considering the fact that $\Gamma(s)$ never attains zero, it implies that $L(f,\ell s+k)$ is non-vanishing in the strip $\frac{1}{2\ell}-\frac{k}{\ell}<\operatorname{Re}(s)<\frac{1-k}{\ell}$ , which in turn implies that $L(f,s)$ does not possess any zero in the strip $\frac{1}{2}<\operatorname{Re}(s)<1$ . From the symmetry of the functional equation, one can conclude that $L(f,s)$ is non-vanishing in $0<\operatorname{Re}(s)<\frac{1}{2}$ . This proves that all the non-trivial zeros of $L(f,s)$ lies on the critical line $\operatorname{Re}(s)=\frac{1}{2}$ . Thus, the GRH follows. This completes the proof of Conjecture 1.3. ∎

Proof of Theorem 3.13.

For the direct part, we assume RH for $\zeta(s)$ to be true and we wish to obtain the bound (3.5) for $P_{k,\sigma,\ell}(x)$ . The proof goes along the similar line of reasoning as in the proof of Theorem 3.10 except for that fact that we start with

\displaystyle\sum_{n\leq x}\sigma_{r}^{-1}(n)=O(x^{\frac{1}{2}+r+\epsilon}),

as established in Lemma 4.7 under the assumption of RH. Thus, we omit the proof here for the direct part.
For the converse part, assuming the validity of the bound (3.5) for $P_{k,\sigma,\ell}(x)$ , that is,

\displaystyle P_{k,\sigma,\ell}(x):=\sum_{n=1}^{\infty}\frac{\sigma_{r}^{-1}(n% )}{n^{k}}\exp\left(-\frac{x}{n^{\ell}}\right)=O_{\epsilon,k,r,\ell}\left(x^{-% \frac{k}{\ell}+\frac{1+2r}{2\ell}+\epsilon}\right),

for any $\epsilon>0$ , we will try to show that all the non-trivial zeros of $\zeta(s)$ are located on the critical line $\operatorname{Re}(s)=\frac{1}{2}$ . Here, we utilize Lemma 4.5 to deduce that

\displaystyle\zeta(k+\ell s)\zeta(k+\ell s-r)\int_{0}^{\infty}x^{-s-1}P_{k,% \sigma,\ell}(x)\mathrm{d}x=\Gamma(-s),

(5.29)

holds within the region $\operatorname{Re}(s)>\frac{1-k+r}{\ell}$ except for $s=0,1,2\cdots$ . Continuing along the path drawn in the proof of Theorem 3.10, one can extend the region of analyticity of (5.29) in $\frac{1}{2\ell}-\frac{k}{\ell}+\frac{r}{\ell}<\operatorname{Re}(s)<\frac{1-k+r% }{\ell}.$ Considering the fact that $\Gamma(s)$ never attains zero, it implies that $\zeta(k+\ell s)$ and $\zeta(k+\ell s-r)$ is non-vanishing in the strip $\frac{1}{2\ell}-\frac{k}{\ell}+\frac{r}{\ell}<\operatorname{Re}(s)<\frac{1-k+r% }{\ell}.$ Since $r$ is a non-negative integer, so $\operatorname{Re}(k+\ell s)>1$ and $\zeta(k+\ell s)$ is non-vanishing in the associated region and as a result, we are unable to derive any useful information. However, the factor $\zeta(k+\ell s-r)$ should also be non-vanishing in the region $\frac{1}{2}<\operatorname{Re}(k+\ell s-r)<1$ , which in turn implies that $\zeta(s)$ does not possess any zeros within the strip $\frac{1}{2}<\operatorname{Re}(s)<1$ and thus the functional equation for $\zeta(s)$ implies that it is non-vanishing in the strip $0<\operatorname{Re}(s)<\frac{1}{2}$ . This proves that all the non-trivial zeros of $\zeta(s)$ must lie on the line $\operatorname{Re}(s)=\frac{1}{2}$ . ∎

6. Numerical evidences for Corollary 3.3 and Theorem 3.5

We used Mathematica software to make the following tables. To obtain this data, we considered the left-hand side sum over $n$ with $200$ terms and right-hand side sum over $n$ with $200000$ terms.

Table 1. Verification of Corollary 3.3

$k$	$x$	Left-hand side	Right-hand side
$2$	$e+1$	$-0.0577422$	$-0.0577021$
$3$	$\pi+1$	$-0.0554663$	$-0.0554081$
$6$	$\pi^{2}$	$-0.0000785321$	$-0.0000787028$
$7$	$e^{3}$	$-8.02103\times 10^{-7}$	$-8.02821\times 10^{-7}$
$11$	$e^{2}+\pi$	$0.000024177$	$0.00002415$

Table 2. Verification of Theorem 3.5

$k$	$r$	$x$	Left-hand side	Right-hand side
$8$	$5$	$e$	$0.0147028$	$0.0147079$
$4$	$1$	$\pi$	$-0.0103086$	$-0.0103073$
$10$	$1$	$\pi+1$	$0.0155115$	$0.0155119$
$11$	$5$	$e+1$	$0.0213461$	$0.0213510$
$15$	$7$	$\pi^{2}$	$0.0000174587$	$0.0000174586$

7. Acknowledgement

The first author’s research is funded by the Prime Minister Research Fellowship, Govt. of India, Grant No. 2101705. The last author wants to thank Science and Engineering Research Board (SERB), India, for giving MATRICS grant (File No. MTR/2022/000545) and SERB CRG grant (File No. CRG/CRG/2023/002122). Both authors sincerely thank IIT Indore for providing conductive research environment.

References

[1] A. Agarwal, M. Garg and B. Maji, Riesz-type criteria for the Riemann hypothesis, Proc. Amer. Math. Soc. 150 (2022), 5151–5163.
[2] T. M. Apostal, Introduction to Analytic Number Theory, Springer, 1998.
[3] L. Báez-Duarte, A sequential Riesz-like criterion for the Riemann hypothesis, Int. J. Math. Math. Sci. 21 (2005), 3527–3537.
[4] S. Banerjee and R. Kumar, Equivalent criterion for the grand Riemann hypothesis associated to Maass cusp forms, Proc. Roy. Soc. Edinburgh Sect. A. (2023). DOI: https://doi.org/10.1017/prm.2023.66
[5] B. C. Berndt, Ramanujan’s Notebooks, Part V, Springer-Verlag, New York, 1998.
[6] R. Bhaskaran, On the versatility of Ramanujan’s ideas, Ramanujan Visiting Lectures, Technical Report 4, Madurai Kamraj University, (1997), 118–129.
[7] K. Chandrasekharan and Raghavan Narasimhan, Hecke’s functional equation and arithmetical identities, Ann. of Math. 74 (1961), 1–23.
[8] J. Cislo and M. Wolf, Criteria equivalent to the Riemann hypothesis, Geometric methods in physics, 268–273, AIP Conf. Proc., 1079, Amer. Inst. Phys., Melville, NY, 2008.
[9] A. Dixit, Character analogues of Ramanujan-type integrals involving the Riemann $\Xi$ -function, Pacific J. Math. 255 (2012), 317–348.
[10] A. Dixit, S. Gupta and A. Vatwani, A modular relation involving non-trivial zeros of the Dedekind zeta function, and the generalized Riemann hypothesis, J. Math. Anal. Appl. 515 (2022), 126435.
[11] A. Dixit, A. Roy, and A. Zaharescu, Ramanujan-Hardy-Littlewood-Riesz phenomena for Hecke forms, J. Math. Anal. Appl. 426 (2015), 594–611.
[12] A. Dixit, A. Roy and A. Zaharescu, Riesz-type criteria and theta transformation analogues, J. Number Theory 160 (2016), 385–408.
[13] P. Epstein, Zur Theorie allgemeiner Zetafunctionen, Math. Ann. 56 (1903), 615–644.
[14] M. Garg and B. Maji, Hardy-Littlewood-Riesz type equivalent criteria for the generalized Riemann hypothesis, Monatsh. für Math. 201 (2023), 771–788.
[15] S. Gupta and A. Vatwani, Riesz type criteria for L-functions in the Selberg class, Canad. J. Math. 76 (2024), 1062–1088.
[16] G. H. Hardy and J. E. Littlewood, Contributions to the theory of the Riemann zeta-function and the theory of the distribution of primes, Acta Math. 41 (1916), 119–196.
[17] A. Ivic, Riemann zeta function, Theory and Applications. Dover Publications, New York, 2003.
[18] H. Iwaniec, E. Kowalski, Analytic Number Theory, American Mathematical Society Colloquium Publications, Vol. 53, 2004.
[19] A. Juyal, B. Maji, S. Satyanarayana, An exact formula for a Lambert series associated to a cusp form and the Möbius function, Ramanujan J. 57 (2022), 769–784.
[20] M. Overholt, A Course in Analytic Number Theory, Graduate studies in mathematics, 160, American Mathematical Society, Rhode Island, 2015.
[21] M. Ram Murty, Problems in Analytic Number Theory, Second Edition, Springer, 2008.
[22] F. W. J. Olver, D. W. Lozier, R. F. Boisvert, C. W. Clark, eds., NIST Handbook of Mathematical Functions, Cambridge University Press, Cambridge, 2010.
[23] R. B. Paris and D. Kaminski, Asymptotics and Mellin–Barnes Integrals, Encyclopedia of Mathematics and Its Applications, vol. 85, Cambridge University Press, Cambridge, 2001.
[24] S. Ramanujan, Notebooks of Ramanujan, Vol 2, Tata Institute of FundamentaI Research, Bombay, 1957.
[25] M. Riesz, Sur l’hypothése de Riemann, Acta Math. 40 (1916), 185–190.
[26] B. Riemann, Ueber die Anzahl der Primzahlen unter einer gegebenen Grösse, Monatsberichte der Berliner Akademie, 1859
[27] E. C. Titchmarsh, The Theory of the Riemann zeta-function, Clarendon Press, Oxford, 1986.
[28] I. Travenec and L. Samaj, Generation of off-critical zeros for hypercubic Epstein zeta functions, Appl. Math. Comput. 413 (2022), 126611.
[29] I. J. Zucker, Exact results for some lattice sums in 2, 4, 6 and 8 dimensions J. Phys. A: Math. Nucl. Gen. 13 (1974), 1568–1575.

	$\displaystyle\left\|\int_{\frac{1}{2}+r+\epsilon+iT}^{1+r+\epsilon+iT}\frac{1}{% \zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w\right\|$	$\displaystyle\ll\int_{\frac{1}{2}+r+\epsilon}^{1+r+\epsilon}x^{\sigma}T^{2% \epsilon-1}d\sigma\ll T^{2\epsilon-1}x^{2+r+\epsilon},$
	$\displaystyle\left\|\int_{\frac{1}{2}+r+\epsilon-iT}^{\frac{1}{2}+r+\epsilon+iT% }\frac{1}{\zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w\right\|$	$\displaystyle\ll\int_{-T}^{T}\|t\|^{2\epsilon}x^{\frac{1}{2}+r+\epsilon}\frac{dt% }{\|t\|}\ll T^{2\epsilon}x^{\frac{1}{2}+r+\epsilon},$
	$\displaystyle\left\|\int_{1+r+\epsilon-iT}^{\frac{1}{2}+r+\epsilon-iT}\frac{1}{% \zeta(w)\zeta(w-r)}\frac{x^{w}}{w}\mathrm{d}w\right\|$	$\displaystyle\ll\int_{1+r+\epsilon}^{\frac{1}{2}+r+\epsilon}x^{\sigma}T^{2% \epsilon-1}d\sigma\ll T^{2\epsilon-1}x^{2+r+\epsilon}.$

Equivalent criteria for the Riemann hypothesis for a general class of L𝐿Litalic_L-functions