Complex cycles on algebraic models of smooth manifolds (Q2267757)

Let \(X\) be a nonsingular real variety, that is, a quasiprojective reduced (but not necessarily irreducible) scheme over \(\mathbb R\), of dimension \(n\). Each algebraic cycle of codimension \(k\) on the complex variety \(X_{\mathbb C}=X\times_{\mathbb R}\mathbb C\) determines a cohomology class in \(H^{2k}(X(\mathbb R),D)\), where \(D\) denotes \(\mathbb Z\) or \(\mathbb Q\). The set \(X(\mathbb C)\) of \(\mathbb C\)-rational points of \(X\) is a complex manifold of complex dimension \(n\) and, if \(X(\mathbb R)\) is nonempty, then it is a real analytic manifold of dimension \(n\). Moreover, \(X_{\mathbb C}\) is a scheme over \(\mathbb C\) whose set \(X_{\mathbb C}(\mathbb C)\) of complex points is identified with \(X(\mathbb C)\). The ring homomorphism \[ \text{cl}_{\mathbb C}:A^*(X_{\mathbb C})=\bigoplus_{k\geq0}A^k(X_{\mathbb C})\to H^{\text{even}}(X(\mathbb C);\mathbb Z)=\bigoplus_{k\geq0}H^{2k}(X(\mathbb C);\mathbb Z) \] is known as the cycle map, and \(\text{cl}_{\mathbb C}(A^k(X_{\mathbb C})\) is the subgroup of \(H^{2k}(X(\mathbb C);\mathbb Z)\) that consists of the cohomology classes corresponding to algebraic cycles on \(X_{\mathbb C}\) of codimension \(k\). The inclusion map \(e:X(\mathbb R)\hookrightarrow X(\mathbb C)\) induces a ring homomorphism \(e^*:H^*(X(\mathbb C);\mathbb Z)\to H^*(X(\mathbb R);\mathbb Z)\), and the subgroup \[ H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z):=e^*(\text{cl}_{\mathbb C}(A^k(X_{\mathbb C})) \] of \(H^{2k}(X(\mathbb R);\mathbb Z)\), and the subring \[ H^{\text{even}}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z):=e^*(\text{cl}_{\mathbb C}(A^*(X_{\mathbb C}))=\bigoplus_{k\geq0}H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z) \] are subtle and important invariants of \(X\). The Chern classes of \(X(\mathbb C)\) are in \(\text{cl}_{\mathbb C}(A^*(X_{\mathbb C}))\), and so \(H^{\text{even}}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z)\) contains the subring of \(H^{\text{even}}(X(\mathbb R);\mathbb Z)\) generated by the Chern classes of the complexification of the tangent bundle to \(X(\mathbb R)\), which can be identified with the restriction to \(X(\mathbb R)\) of the tangent bundle to \(X(\mathbb C)\). A challenging problem is to decide what cohomology classes, if any, belong to \(H^{\text{even}}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z)\), and this depends on the algebro-geometric properties of \(X\). In general, the groups \(H^{2k}_{\mathbb C-\text{alg}}(-;\mathbb Z)\) are very difficult to handle even for \(k=1\), and they become more manageable after tensoring with \(\mathbb Q\). The image of \(H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z)\otimes_{\mathbb Z}\mathbb Q\) under the canonical isomorphism \[ H^*(X(\mathbb R),\mathbb Z)\otimes_{\mathbb Z}\mathbb Q\to H^*(X(\mathbb R),\mathbb Q) \] is denoted by \(H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q)\). By construction, \[ H^{\text{even}}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q):=\bigoplus_{k\geq0}H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q) \] is a \(\mathbb Q\)-subalgebra of \(H^{\text{even}}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q)\). It contains the rational Pontryagin classes of \(X(\mathbb R)\), and the main results of the paper under review concern \(H^{2k}_{\mathbb C-\text{alg}}(-;\mathbb Q)\) and \(H^{\text{even}}_{\mathbb C-\text{alg}}(-;\mathbb Q)\). By \textit{A. Tognoli}'s theorem [Ann. Sc. Norm. Super. Pisa, Sci. Fis. Mat., III. Ser. 27, 167--185 (1973; Zbl 0263.57011)], every compact smooth manifold \(M\) is diffeomorphic to the set \(X(\mathbb R)\) of real points of a nonsingular projective real algebraic variety \(X\), which is called an algebraic model of \(M\). Indeed \(X(\mathbb R)\) is a dense subset of \(X\) and this implies the irreducibility of \(X\), provided \(M\) is connected. In this paper the author investigates how the \(\mathbb Q\)-vector space \(H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q)\) varies as \(X\) runs through the class of all algebraic models of \(M\). The \(i\)-th rational Pontryagin class of \(M\) is denoted by \(p_i(M)\in H^{4i}(M,\mathbb Q)\), and one of the more important results of the paper is the following theorem, which under mild assumptions improves an old result by \textit{J. Bochnak} and the author [Proc. Am. Math. Soc. 114, No. 4, 1097--1104 (1992; Zbl 0760.57011)]. Theorem 1. Let \(M\) be a compact connected orientable smooth manifold. For any nonnegative integer \(k\) and any \(\mathbb Q\)-vector subspace \(A\) of \(H^{2k}(M;\mathbb Q)\), the following conditions are equivalent: (1) There exist an algebraic model \(X\) of \(M\) and a smooth diffeomorphism \(\varphi: X(\mathbb R)\to M\) such that \(H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q)=\varphi^*(A)\). (2) The cup product \(p_{i_1}(M)\cup\cdots\cup p_{i_r}(M)\) belongs to \(A\) for all nonnegative integers \(i_1,\dots,i_r\) satisfying \(2(i_1+\cdots+i_r)=k\). Recall that a compact oriented smooth manifold is said to be an oriented boundary if it is the boundary, endowed with the induced orientation, of a compact oriented smooth manifold with boundary. As a consequence of Theorem 1 the author improves one of the main results in [\textit{J. Bochnak} and \textit{W. Kucharz}, Ann. Math. (2) 128, No. 2, 415--433 (1988; Zbl 0674.14013)]. The statement is the following: Corollary 1. Let \(M\) be a compact connected oriented smooth manifold of dimension \(2k\). Then, the following conditions are equivalent: (1) There exists an algebraic model \(X\) of \(M\) with \(H^{2k}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Z)=0\). (2) The disjoint union of two copies of \(M\) is an oriented boundary. In fact, both conditions are satisfied if \(k\) is odd. The author states also the following Conjecture, which generalizes the previous Theorem 1. Conjecture. Let \(M\) be a compact connected orientable smooth manifold. For any \(\mathbb Q\)-subalgebra \(A\) of \(H^{\text{even}}(M;\mathbb Q)\), the following conditions are equivalent: (1) There exists an algebraic model \(X\) of \(M\) and a smooth diffeomorphism \(\varphi: X(\mathbb R)\to M\) such that \(H^{\text{even}}_{\mathbb C-\text{alg}}(X(\mathbb R);\mathbb Q)=\varphi^*(A)\). (2) The rational Pontryagin classes of \(M\) are in \(A\). Indeed, it follows readily from Theorem 1 that this conjecture is true if either \(\dim(M)\leq3\) or \(\dim(M)=4\) and \(M\) is not the boundary of a compact orientable smooth manifold with boundary (in the latter case \(H^4(M;\mathbb Q)\) is generated by \(p_1(M))\). In arbitrary dimension the author proves the conjecture only when \(A\) is generated by the rational Pontryagin classes of \(M\). I believe that this article is a masterpiece of very beautiful mathematics, written in the penetrant and transparent style of the author.