Reciprocity laws for simple algebras over function fields of number curves (Q4953285)

The main object of the paper under review is a central simple algebra defined over a function field \(F\) of transcendence degree one over a ground field \(k\) of characteristic zero. In other words, \(F=k(X)\) is the field of rational functions of a smooth geometrically connected projective curve \(X\) defined over \(k\). Roughly speaking, the authors' goal is to understand to what extent a given algebra is determined by its local invariants, or, in other words, to describe the reciprocity laws for local invariants. To be more precise, one can write down a complex of abelian groups \[ 0\to\text{{Br }}X\to\text{{Br }}k(X)\overset{\text{{inv}}}{}\bigoplus_{x\in X^1}H^1(k(x),\mathbb Q/\mathbb Z)\overset{\Sigma }{} H^1(k,\mathbb Q/\mathbb Z), \] where \(\text{{inv}}\) is the direct sum of local invariants and \(\Sigma\) is the sum of corestriction maps. If \(X={\mathbb P}^1\), it is known (Faddeev-Grothendieck) that this complex is exact and, moreover, \(\Sigma\) is surjective (local invariants of a central simple algebra over \(k(t)\) satisfy the Faddeev reciprocity law, and, conversely, any system satisfying this reciprocity law comes from some central simple algebra). If \(X\) is a curve of positive genus, this is no longer true: in general, \(\text{{im}}(\text{{inv}})\neq\text{{ker}}(\Sigma)\). The authors study the quotient \(\Delta =\text{{ker}}(\Sigma)/\text{{im}}(\text{{inv}})\) which can be thought of as the obstruction group for realizing characters satisfying the reciprocity law as systems of local invariants of algebras over \(k(X)\). The authors focus on the case where \(k\) is a number field and \(X\) has a \(k\)-point. \textit{B.~Bekker} [Mat. Zametki 17, 419-422 (1975; Zbl 0321.14015)] proved that under these hypotheses \(\Delta\) is a 2-group (trivial if \(k\) is totally imaginary). The main result of the paper under review makes this more precise: \(\Delta\cong ({\mathbb Z}/2)^{r_1+\dots +r_s-s}\); here \(s\) is the number of real embeddings \(\rho _i\colon k\to\mathbb R\) and \(r_i\) is the number of components of \(X_i(\mathbb R)\) in the real topology where \(X_i=X\times _k\mathbb R\) with the product taken with respect to \(\rho _i\).