On the integral Hodge conjecture for real varieties. I (Q2006705)

Let \(X\) be a proper smooth complex algebraic variety. The Hodge conjecture claims that the Hodge classes \(\mbox{Hdg}^{2k}(X, \mathbb Z) \subset H^{2k}(X, \mathbb Z)\) coincide with the cycle classes given by algebraic subvarieties \(H^{2k}_{\mathrm{alg}}(X, \mathbb Z) \subset H^{2k}(X, \mathbb Z)\) up to tensoring with \(\mathbb Q\). One says that the variety \(X\) satisfies the {\em integral} Hodge conjecture if the inclusion \(H^{2k}_{\mathrm{alg}}(X, \mathbb Z) \subset \mbox{Hdg}^{2k}(X, \mathbb Z) \) is an equality. There are now many examples of complex varieties that do not satisfy the integral Hodge conjecture, but \textit{C. Voisin} showed in [Adv. Stud. Pure Math. 45, 43--73 (2006; Zbl 1118.14011)] that it holds for uniruled and Calabi-Yau threefolds. \newline In this paper the authors introduce an analogue of the integral Hodge conjecture in the real setting: let \(X\) be a smooth proper real algebraic variety. Borel and Haefliger associated to every algebraic cycle of codimension \(k\) a cycle class with coefficients in \(\mathbb Z/2\mathbb Z\), so we get a subgroup \[ H^{k}_{\mathrm{alg}}(X(\mathbb R),\mathbb Z/2\mathbb Z) \subseteq H^{k}(X(\mathbb R),\mathbb Z/2\mathbb Z). \] The authors define a subgroup \[ \mbox{Hdg}^{2k}_G(X(\mathbb C),\mathbb Z(k))_0 \subseteq H^{2k}_G(X(\mathbb C),\mathbb Z(k)) \] where \(G=\mbox{Gal}(\mathbb C/\mathbb R)\) acts on the space \(X(\mathbb C)\) and on the group \(\mathbb Z(k)=(\sqrt{-1})^k\mathbb Z\) and where \(H^{2k}_G(X(\mathbb C),\mathbb Z(k))\) denotes \(G\)-equivariant cohomology in the sense of Borel. One says that \(X\) satisfies the real integral Hodge conjecture if every element of this subgroup is the equivariant cycle class of some algebraic cycle of codimension \(k\). \newline In a first step the authors consider the validity of the real integral Hodge conjecture in a setting that is analogous to Voisin's statement: let \(X\) be a smooth and proper real threefold. Assume that the corresponding complex threefold is rationally connected or is simply connected Calabi--Yau. Then the real integral Hodge conjecture holds for \(X\) if and only if \[ H^2_{\mathrm{alg}}(X(\mathbb R),\mathbb Z/2\mathbb Z)=H^2(X(\mathbb R),\mathbb Z/2\mathbb Z), \] if \(X(\mathbb R)\neq\emptyset\), or if there exists a geometrically irreducible curve of even geometric genus on \(X\), if \(X(\mathbb R)=\emptyset\). \newline In a second step the authors establish some consequences of the real integral Hodge conjecture for 1-cycles: let \(X\) be a smooth, proper and geometrically irreducible real variety, of dimension \(d\geq 1\) such that that \(H^2(X,\mathcal O_X)=0\). Assume that the real integral Hodge conjecture for \(1\)-cycles holds for \(X\). Then the following holds: \begin{itemize} \item The subgroup \[ H^{d-1}_{\mathrm{alg}}(X(\mathbb R),\mathbb Z/2\mathbb Z) \subseteq H^{d-1}(X(\mathbb R),\mathbb Z/2\mathbb Z) \] is the exact orthogonal complement, under the Poincaré duality pairing, of the image of \(\mbox{Pic}(X)[2^\infty]\) by the Borel-Haefliger cycle class map \(\mbox{Pic}(X) \to H^1(X(\mathbb Z),\mathbb Z/2\mathbb Z)\). \item There exists a geometrically irreducible curve of even geometric genus in~\(X\) if and only if the natural map \(\mbox{Pic}(X)[2^{\infty}]\to \mbox{Pic}(X_\mathbb C)^G[2^{\infty}]\) is onto. \end{itemize} If one assumes furthermore that \(\mbox{CH}_0(X_\mathbb C)\) is supported on a surface (e.g. if the corresponding complex threefold is uniruled), the real Hodge conjecture for \(1\)-cycles implies that the image of \(\mbox{CH}_1(X)[2^\infty]\) by the Borel-Haefliger cycle class map \[ \mbox{CH}_1(X) \to H^{d-1}(X(\mathbb R),\mathbb Z/2\mathbb Z) \] is the exact orthogonal complement, under the Poincaré duality pairing, of the subgroup \(H^1_{\mathrm{alg}}(X(\mathbb R),\mathbb Z/2\mathbb Z) \subseteq H^1(X(\mathbb R),\mathbb Z/2\mathbb Z)\). \newline We refer to the very informative introduction for a detailed description of the techniques appearing in the proofs of these statements. Note also that in the sequel to this paper [J. Éc. Polytech., Math. 7, 373--429 (2020; Zbl 1439.14037)] the authors verify the real integral Hodge conjecture for several classes of varieties.