A summation formula in algebraic number fields and applications. I (Q1813406)

Let \(K\) be an algebraic number field of degree \(n=[K:\mathbb{Q}]=r_ 1+2r_ 2\), \(r_ 1\), \(2r_ 2\) denoting the number of real, complex conjugates, resp. Let \(E\) denote the group of units of the integer ring \({\mathcal O}_ K\) of \(K\), so that the order of \(E\) is \(r_ 1+r_ 2-1=r\), say. For an arithmetical function \(f\) possessing the invariance property \(f(\eta\nu)=f(\nu)\), \(0\neq\nu\in{\mathcal O}_ K\), where \(\eta\) runs through a free subgroup \(U\) of \(E\) consisting of totally positive units, Siegel's summation formula refers to the evaluation of sums \[ F(x)=\sum_ \nu f(\nu)\Phi(|\nu^{(1)}| x_ 1^{- 1},\dots,|\nu^{(r+1)}| x^{-1}_{r+1}), \] where \(\nu\) runs through totally positive integers (i.e. those integers for which \(\nu^{(p)}>0\), \(p=1,\dots,r_ 1)\) and \(\Phi\) denotes the characteristic function of the unit interval \((0,1]^{r+1}\) and \(x_ p\) are large parameters. Siegel's summation formula has been very useful as a substitute for Perron's formula in algebraic number fields, and has been successively applied, e.g. by \textit{W. Grotz} [Acta Arith. 38, 69-95 (1980; Zbl 0363.12011); Monatsh. Math. 88, 219-228 (1979; Zbl 0386.12009)]. The author provides a new summation formula, related to but more general than, Siegel's and gives improved generalizations of Grotz's results, notably on Piltz's divisor problem, sum-of-divisors function, etc. For a generalized Größencharacter \(\Lambda=\Lambda(\nu)\) and a sign character \(v=v(\nu)\), and a fairly general weight function \(\Phi:\mathbb{R}_ +^{r+1}\to\mathbb{C}\), this summation formula handles the series of the form \[ G(x)=G(x;\Lambda vf;\Phi)=\sum_{\nu\neq 0}\Lambda(\nu)v(\nu)f(\nu)\Phi(|\nu^{(1)}| x_ 1,\dots,|\nu^{(r+1)} | x_{r+1}) \] and gives (Theorem 2.2) an expression as a series of absolutely convergent integrals (via Fourier expansion ``with respect to units'') for the integral average \(J_ \varepsilon(G)\) of \(G\): \[ J_ \varepsilon(G(x))=(4\pi\varepsilon)^{- (r+1)/2}\int_{\mathbb{R}^{r+1}}G(x_ 1e^{v_ 1 },\dots,x_{r+1}e^{v_{r+1}})\exp\left(-{1\over 4\varepsilon}\sum^{r+1}_{p=1} | v_ p|^ 2\right)^{1/2}dv_ 1\dots dv_{r+1}. \] Through a Tauberian process (Theorem 3.1) one can infer back from \(J_ \varepsilon(G)\) to \(G\) and (choosing \(\Phi\) suitably) one obtains asymptotic formulas for the original sums. The author promises to give a general summation formula in which \(\Phi\) depends on conjugates themselves in Part II (to appear).