Periods of Hodge structures and special values of the gamma function (Q522198)

Motivated by his proof of the Chowla-Selberg formula, \textit{B. H. Gross} made a conjecture (whose precise formulation was suggested by Deligne) which relates special values of the gamma function and periods of ``motives with abelian complex multiplication'' [Invent. Math. 45, 193--211 (1978; Zbl 0418.14023)]. By definition these are motives equipped with an action of an abelian CM number field of degree (over \(\mathbb{Q}\)) equal to the rank of the motive. Examples arising from geometry can be constructed as follows. Let \(X\) be a variety and \(g\) an automorphism of \(X\) of order \(n\) and consider the motive \(H^i(X)\) for a fixed cohomological degree \(i\). Write \(H^i(X,g)\) for the biggest direct factor where the eigenvalues for the action of \(g\) are all primitive \(n\)-th roots of unity. Notice that \(H^i(X,g)\) is a \(\mathbb{Q}(\xi_n)\)-vector space, where \(\mathbb{Q}(\xi_n)\) is the \(n\)-th cyclotomic field. Hence, the determinant \( \det_{\mathbb{Q}(\xi_n)} H^i(X,g)\) is a motive with complex multiplication by the abelian number field \(\mathbb{Q}(\xi_n)\). Notice that \(det_{\mathbb{Q}(\xi_n)} H^i(X,g)\) can be seen as a direct factor of the cohomology of a power of \(X\) by the Künneth formula. The conjecture is known in degree \(i=1\) by a result of \textit{G. W. Anderson} [Compos. Math. 45, 315--332 (1982; Zbl 0501.14025)] but little is known in higher degree. The paper under review studies periods of (the product of) these geometric motives \( \otimes_i \det_{\mathbb{Q}(\xi_n)} H^i(X,g)\) constructed above. In particular it shows that the alternating variant of Gross-Deligne conjecture holds true. Alternating variant means that one considers the alternating product of the periods living in different cohomological degrees. In general it seems very hard with the present technology to isolate the periods of a single cohomological degree from this alternating product. Examples where this is still tractable are given in the paper (e.g. the case of hypersurfaces, see Corollary 2.1). The author also states a result (Theorem C) which is a positive characteristic analogue of his main result, where periods are replaced by epsilon factors. Its proof will appear in a sequel of the paper. The main result of the paper is an equality between periods and special values up to an algebraic constant \(\lambda\). Such an equality lives in \(\mathbb{C}^*/ \overline{\mathbb{Q}}^*\). If one applies the complex norm to this equality one obtains an equality in \(\mathbb{R}^*/ (\overline{\mathbb{Q}}\cap\mathbb{R})^* \) which was shown by \textit{V. Maillot} and \textit{D. Roessler} in the case when \(n\) is prime [Ann. Math. (2) 160, No. 2, 727--754 (2004; Zbl 1077.14032)] (see also \textit{C. Soulé} [in: Séminaire Bourbaki. Volume 2005/2006. Exposés Nos. 952--966. Paris: Société Mathématique de France. 75--98, Exp. No. 955, (2007; Zbl 1197.14022)]). Moreover, the author can give a control on the field \(K \subset \overline{\mathbb{Q}}\) where the algebraic constant \(\lambda\) lives (Theorem A). When \(X\) is a curve, the result of the paper recovers a result of \textit{T. Terasoma} [J. Reine Angew. Math. 433, 143--159 (1992; Zbl 0753.14022)] (see Theorem B). The idea of the proof (which the author attributes to Bloch) is to transform this absolute problem into a relative one by considering (a resolution of) the \(g\)-invariant quotient map \(f: X \rightarrow X/(g).\) Then absolute periods can be related to periods of vector bundles with connection on the base. This strategy is very clearly explained in the introduction, where the author gives the example of a Fermat curve covering \(\mathbb{P}^1\). In that case the computation of periods gives back Euler's formula relating the beta and the gamma functions. In order to compute periods of connections a very general result is available, namely a formula of \textit{T. Saito} and \textit{T. Terasoma} [Proc. Japan Acad., Ser. A 69, No. 5, 131--135 (1993; Zbl 0824.19003)]. One of the contributions of the paper is to shed a light on this powerful tool. A self-contained account is given, together with examples. Computing the different terms in the formula is a rather non-trivial task. Already the periods of the unit object asks a certain amount of work (Section 3.7). These computations will certainly be useful in other contexts.