Retracted: A note on number knots and the splitting of the Hilbert class field (Q1988604)

Let \(L\) be a number field, \(L^{\ast }\) its multiplicative group, \(L_{\mathrm{sep}}\) a separable closure of \(L\), \(J_L\), \(C_L\), \(I_L\), \(P_L\) and Cl\(_L\) the idele group, the idele class group, the group of fractional ideals, the group of principal ideals and the ideal class group of \(K\), respectively. Denote by \(E_L\) the global unit group, by \(U_L\) the unit idele group, and by \(\mathcal{E}_L\) the the idele unit class group, and let \(H_L\) be the Hilbert class field of \(L\), i.e. the maximal unramified abelian extension of \(L\) in \(L_{\mathrm{sep}}\)). It is well-known that the Galois group \(\mathcal{G}(H_L/L)\) is isomorphic to Cl\(_L\). Suppose further that \(K/k\) is a Galois extension of number fields, put \(G = \mathcal{G}(K/k)\), and denote by \(N\) the relative norm \(N = N_{K/k}\). Then \(H_{K/k}\) is a Galois extension and the field tower of Galois extensions \(H_{K/K/k}\) defines a group extension with (abelian) kernel \(\mathcal{G}(H_{K/K})\) and factor set \(\alpha\): \[ \alpha \colon 1 \to \mathcal{G}(H_{K/K}) \to \mathcal{G}(H_{K/k}) \to G \to 1.\tag{1} \] Number knots are certain groups associated with the extension \(K/k\), which play a role in the problem of whether the group extension (1) splits. Jehne has exhibited the following five number knots: the first unit knot \(\omega_{K/k} = E_k \cap NK^{\ast }/NE_K\), the second unit knot \(\omega '_{K/k} = E_k \cap NU_K/NE_K\), Scholz's number knot \(\nu_{K/k} = k^{\ast } \cap NJ_K/NK^{\ast }\), the divisor knot \(\delta_{K/k} = P_k \cap NI_K/NP_K\), and the idele class knot \(\gamma_{K/k} = \mathcal{E}_k \cap NC_K/N\mathcal{E}_K\) (see [\textit{W. Jehne}, J. Reine Angew. Math. 311--312, 215--254 (1979; Zbl 0432.12006)]). The study of knots in number fields is motivated by their expected (or confirmed) rich arithmetic content. For example, by a theorem due to Scholz, the knot \(\nu_{K/k}\) is trivial if and only if the Hasse Norm Principle holds for \(K/k\); this basically formulates in non-cohomological terms Tate's theorem on the obstruction to the validity of the Hasse Norm Theorem (see Sect. 11.4 of [\textit{J. T. Tate}, in: Algebraic number theory. 162--203 (1967; Zbl 1179.11041)]. The interest in the arithmetic content of knots is also due to the fact that many of them can be viewed as Galois groups of certain subfields of \(H_K\) (see Theorem 3 of [Jehne, loc. cit.]. For example, there are group isomorphisms \(\delta_{K/k} \cong \mathcal{G}(H_K^{\ast }/H_kK)\) and \(\gamma_{K/k} \cong \mathcal{G}(H_K^{\prime }/H_kK)\), where \(H_K^{\prime }\) is the abelian genus field and \(H_K^{\ast }\) is the central genus field of \(K\) over \(k\). The paper under review introduces further cohomological knots \(\nu_n(K/k)\), for \(n = -1; 0; 2\) and \(\tau_{K/k}\), and relates them with the number knots of Scholz and Jehne. It characterizes the splitting condition on (1) in terms of triviality of certain cohomological knots. The author defines a global inertia group, and proves that it coincides with \(G\) in case the degree \([K\colon k]\) equals the least common multiple of the ramification indices for \(K/k\) (finite or infinite). This main result of the reviewed paper shows that, for \(G\) abelian, the group extension (1) splits whenever \(H_k \cap K = k\). When \(G\) is cyclic, the stated result has earlier been obtained by \textit{B. F. Wyman} [Scripta Math. 29, 141--149 (1973; Zbl 0273.12007)] and \textit{R. Gold} [Ill. J. Math. 21, 66--69 (1977; Zbl 0346.12004)]. In the final section, the author introduces a group \(J\) which can be identified with a subgroup of Ker\((j)\), where \(j\colon\mathrm{Cl}_k \to\mathrm{Cl}_K\) is the extension map of ideals. He shows that if the Scholz number knot is isomorphic to \(H^{-1}(C_K) \cong H_2(G, \mathbb{Z})\), the Pontryagin dual of the Schur multiplier of \(G\), then: (a) the splitting of (1) implies \(\nu_{-1} = 1\) and \(H^{-1}(\mathrm{Cl}_K) \cong J\); (b) if \(G\) is nilpotent, \(\nu_{-1} = 1\) and \(J = 1\), then (1) splits. This theorem covers several results obtained by \textit{R. J. Bond} [J. Number Theory 42, No. 3, 349--360 (1992; Zbl 0778.11063)].