The spin of prime ideals (Q376215)

The distribution of prime ideals in number fields usually involves a combination of techniques from algebraic and analytic number theory involving \(L\)-functions and their Euler product representation. In situations where an Euler product is not available, as it is in this paper, a different approach is needed. Let \(K\) be a number field and a fixed nontrivial automorphism \(\sigma:K\to K\). The authors associate to ideals in \(K\) an invariant called the spin for which the associated \(L\)-function does not possess Euler products. The authors restrict themselves to a special class of number fields, where the distribution of the spin over primes can be handled by using the well-known technique of transforming sums over primes to special congruence sums and bilinear forms. Their results in full generality are dependent on a conjecture. Let \(K/\mathbb{Q}\) be a Galois extension of degree \(n\geq 3\) with cyclic Galois group \(G=\mathrm{Gal}(K/\mathbb{Q})=\langle \sigma \rangle\). Furthermore assume that \(K\) is totally real and that the totally positive units are exactly the squares of units. Two infinite classes of number fields that satisfy these conditions are described. Namely, such a family of cyclic cubic number fields has been studied by \textit{D. Shanks} [Math. Comput. 28, 1137--1152 (1974; Zbl 0307.12005)], and such a family of cyclic quintic number fields has been constructed by \textit{E. Lehmer} [Math. Comput. 50, No. 182, 535--541 (1988; Zbl 0652.12004)] . If \(\mathfrak{a}\) is a principal ideal coprime to \(2\) and \(\alpha\) is a totally positive generator of \(\mathfrak{a}\), the spin of \(\mathfrak{a}\) is defined by \[ \mathrm{spin}(\sigma,\mathfrak{a}):=\left(\,\frac{\alpha}{\mathfrak{a}^{\sigma}}\,\right) \,, \] where \(\left(\,\frac{\alpha}{\mathfrak{b}}\,\right)\) denotes the quadratic residue symbol in \(K\). Note that if \(\mathfrak{a}\) and \(\mathfrak{a}^{\sigma}\) are coprime then \(\mathrm{spin}(\sigma,\mathfrak{a})\in\{-1,1\}\). Although the quadratic residue symbol \(\left(\,\frac{\alpha}{\mathfrak{b}}\,\right)\) has multiplicative properties, the spin does not behave in a purely multiplicative fashion, so that techniques of \(L\)-functions do not apply. The main results of the authors are based on the assumption that the following Conjecture \(C_n\) holds: Let \(n\geq 3\), \(Q\geq 3\), \(N\leq Q^{\frac{1}{n}}\), \(q\leq Q\), and let \(\chi\) be any real non-principal Dirichlet character mod \(q\). The real character sums \[ S_{\chi}(M,N):=\sum_{M<n\leq M+N} \chi(n),\quad N\geq 1\,, \] over (short) intervals satisfy the condition that \[ S_{\chi}(M,N) \ll Q^{\frac{1-\delta}{n}+\varepsilon}\,, \] with some \(\delta=\delta(n)>0\) and any \(\varepsilon>0\), and the implied constant depending only on \(\varepsilon\) and \(n\). For instance, due to a result of \textit{D. A. Burgess} [Proc. Lond. Math. Soc. (3) 13, 524--536 (1963; Zbl 0123.04404)], \(C_3\) holds with \(\delta(3)=\frac{1}{48}\). Let \(n\geq 3\) and assume the above mentioned Conjecture \(C_n\) with exponent \(\delta\leq \frac{2}{n}\). The first main theorem of the authors states that \[ \sum_{\substack{\mathfrak{p}\mathrm{ principal} ,\\ \mathrm {N}(\mathfrak{p})\leq x}} \ll x^{1-\nu+\varepsilon}\,, \] where \(\mathfrak{p}\) runs over prime ideals coprime to \(2\), \(\nu=\nu(n)=\frac{\delta}{2n(12n+1)}\), and the implied constant depends only on \(\varepsilon\) and the number field \(K\). The second main theorem states that the bound in the first theorem still holds when the sum is further restricted to principal prime ideals which have a totally positive generator \(\pi\) satisfying \(\pi\equiv \mu \pmod{\mathfrak{M}}\), where \(\mathfrak{M}\) is an integral ideal of \(K\) divisible by \(2\) and \(\mu\) is an integer of \(K\) coprime to \(\mathfrak{M}\). The main results of this paper have implications for the distribution of the Selmer rank on elliptic curves. Let \(E\) be an elliptic curve over \(\mathbb{Q}\) and assume that the field \(K=\mathbb{Q}(E[2])\) satisfies the necessary conditions in this paper with \(n=3\) and generator \(\sigma\) of the cyclic Galois group of \(K\). Let \(p\) be any rational prime that splits completely in \(\mathbb{Q}(E[4])\), and let \(\mathfrak{p}\) be any of the primes of \(K\) above \(p\). Suppose further that \(\mathfrak{p}\) has a totally positive generator \(\pi\) such that \(\pi\) is a square in \(K_{\ell}\) for every prime \(\ell\) such that \(\ell=2\) or \(E\) has bad reduction at \(\ell\) and \(\ell\) is unramified in \(K\). Then viewing \(\mathrm {Sel}_2(E)\) and \(\mathrm {Sel}_2(E^{(p)})\) as subgroups of \(H^1(\mathbb{Q},E[2])\), one identifies \(\mathrm{Sel}_2(E) \subseteq \mathrm {Sel}_2(E^{(p)})\), and one obtains: If \(\mathrm {spin}(\sigma,\mathfrak{p})=+1\), then \(\dim_{\mathbb{F}_2} \mathrm{Sel}_2(E^{(p)})=\dim_{\mathbb{F}_2} \mathrm {Sel}_2(E)+2\), and if \(\mathrm{spin}(\sigma,\mathfrak{p})=-1\), then \(\dim_{\mathbb{F}_2} \mathrm{Sel}_2(E^{(p)})=\dim_{\mathbb{F}_2} \mathrm{Sel}_2(E)\).