A number field extension of Ramanujan’s identity and its integral representation via the generalized Riemann \Xi-function

[] Let us define

Considering the function \phi(s) as

we have f(t) = \phi(it)\phi(-it). Substituting this into (?) with y=e^{2z} and applying the functional equation s\Gamma(s) = \Gamma(s+1), we obtain

where in the penultimate step, we invoked (?), followed by Legendre’s duplication formula (?), Euler’s reflection formula (?) for \Gamma(s), and the asymmetric functional equation (3.1) for \zeta_\mathbb{F}(s). We now denote the integrand in (?) by

To evaluate the integral in (?), we shift the line of integration from \operatorname{Re}(s)=\frac{1}{2} to \operatorname{Re}(s)=1+\delta, with \delta\in(0,1). This shift introduces a pole of order r_2+2 at s=1, which must be treated carefully. Given T>0, the residue theorem implies

where

From the Stirling formula of gamma function, we deduce that the integrals along the horizontal lines from \frac{1}{2}\pm iT to 1\pm \delta+iT vanish as T \rightarrow \infty. Now we further simplify the vertical integral from 1+\delta-i\infty to 1+\delta+i\infty using the series definition (?) of \zeta_\mathbb{F}(s) and interchange the integral with the summation due the absolute convergence to have

where

Hence substituting the value of (?) in (?) with T \to \infty and using the final value of (?) in (?), we get

where

In (?), we have used the fact that \overline{\Gamma(s)} = \Gamma(\overline{s}) and \Xi_{\mathbb F}(t) is real for real values of t. Substituting z=\frac{1}{2}\log\alpha into (?) and observing that y=e^{2z}=\alpha with \alpha\beta=1, we get

Since the left hand side of the above equation is unchanged under the substitution \alpha \mapsto \beta, we interchange \beta by \alpha in (?) and combine it with (?) to get the desired transformation formula (2.1).

[] Taking \mathbb{F} = \mathbb{Q} in Theorem 2.1, and using (?) together with the fact that \mathtt{a}_{\mathbb{Q}}(m) = 1, we obtain

Now, we calculate the term \mathcal{R}^{(r_2+2)}_1\left(\frac{1}{x}\right) = \mathcal{R}^{(2)}_1\left(\frac{1}{x}\right) in (2.1), where

We use the following Laurent series expansions to find the above residual term.

Hence, collecting the coefficient of (s-1) in the Laurent series expansion of (s-1)^{2}\frac{\zeta(s)\zeta(1-s)}{\sin(\pi s)}x^s, we have

Hence, substituting the value of (Proof) and (Proof) in Theorem 2.1, we get the desired identity (?).

[] Let us define

We know that f(t) = \phi(it)\phi(-it), so we have

where \xi_{\mathbb{F}}(s) is defined in (?). Substituting the above value of \phi(s) in (?) with y=e^{2z}, we obtain

where in the penultimate step, we invoked (?). To solve the above integral in (Proof), we shift the line of integration from \operatorname{Re}(s)=\frac{1}{2} to \operatorname{Re}(s)=1+\delta, with \delta\in(0,1). This shift introduces a pole of order k at s=1, which must be treated carefully. Let T>0, the residue theorem implies

where

The horizontal integrals vanish as T \rightarrow \infty due to the Stirling formula for the gamma factors in the integrand and we are left with

Note that the integral on the right hand side has been evaluated by the author and Maji [BM2, p. 14], and can be written as

where S_{\mathbb{F}, k}(x) is given by

Substituting (Proof) in (Proof) and then using the final value of the integral on the left hand side of (Proof) into (Proof), we get

Substituting z=\frac{1}{2}\log\alpha into the last equation and observing that y=e^{2z}=\alpha with \alpha\beta=1, we get

Since the left-hand side of the above equation is invariant under the interchange \alpha\leftrightarrow\beta, we replace \beta by \alpha in (Proof) and combine the resulting relation with (Proof). We also shuffle the residual terms and use \alpha = \frac{1}{\beta} to obtain

which is the desired identity (2.3).

[] For k=1, the term R^{(k)}_1(x) in Theorem 2.3 reduces to

Substituting this expression for R^{(1)}_1(x) with k=1 into Theorem 2.3 yields the desired identity in Corollary 2.4.

[] For k=2, the term R^{(k)}_1(x) in Theorem 2.3 reduces to

To solve the above term, we use (?) and (?) to find the Laurent series expansion of \Omega_\mathbb{F}^2(s)x^s around 1 to have

In particular,

Upon inserting the above expression for R^{(2)}_1(x) into (2.3) and taking k=2 in Theorem 2.3, we obtain Corollary 2.6.

[]

For \mathbb{F}=\mathbb{Q}, we have

Moreover, by (?) and (?),

Consequently, the function \mathcal{H}_{\mathbb{F},2}(x) in Corollary 2.6 becomes

Substituting the above expression for \mathcal{H}_{\mathbb{Q},2}(x) into (2.6) and taking \alpha\beta=1, we obtain

Using \beta=1/\alpha and rearranging the terms, the preceding identity takes the form

This is precisely the extended Ramanujan–Koshliakov formula (?).

From (?), we have

We use the reflection formula (?) in (?) to obtain

We now employ (?) in the last equation to get

which gives the desired result.

We consider a rectangular contour with vertices 1+\delta - iT, 1+\delta + iT, \alpha +iT and \alpha -iT, where \alpha < 0. By Cauchy’s residue theorem, we have

where \mathcal{R} is the sum of the residues of the integrand function inside the contour. As T \to \infty, the horizontal integrals vanish because the exponential decay of \sin^{-d}(\pi s) dominates the polynomial growth of \zeta_{\mathbb F}(1-s). Taking \alpha \to -\infty, the left vertical integral likewise vanishes, as \sin^{-d}(\pi s) continues to suppress \zeta_{\mathbb F}(1-s) while y^{-s} decays to zero under the absolute convergence of \zeta_{\mathbb F}(1-s). We now analyse the poles of the integrand function. The Dedekind zeta function \zeta_{\mathbb F}(1-s) has a zero of order r_1+r_2-1 at s=0 and simple pole at s=1. Thus, the integrand has a pole of order d+1 at s=0, a pole of order r_2+1 at s=1, and poles of order d at the negative integers. Therefore,

where the terms R^{(d+1)}_0(y), R^{(r_2+1)}_1(y) and R^{(d)}_{-k}(y) are given by