A remark on the generalized Hodge conjecture (Q463955)

Give a smooth complex projective variety \(X\) of dimension \(n\), the leading actor in Grothendieck's generalized Hodge conjecture is the rational subspace \(N_m H_{r}\) inside \(H_{r}(X, \mathbb Q)\) which is generated by the images of \(H_{r}(W, \mathbb Q)\), where \(W \to X\) varies in the family of algebraic varieties \(W\) of dimension at most \(n-m\). Let \(N_{m ,r}:= {N_m H_{r}} \otimes {\mathbb C} \) be the complex span. The main result of the present paper is an interesting and somewhat unexpected characterization of \(N_{m ,r}\). The author's hope is that his work might turn out to be of help for a deeper understanding of the conjecture. An algebraic set admits a triangulation. On \(X\) a convenient system of open sets and holomorphic local coordinates on them is chosen. The first basic fact proved here is that one can use singular \(r\)-simplexes \(\sigma_h:\bar\Delta_r \to X\) with the property that (i) the range of \(\sigma_h\) lies in one of the open sets (ii) the restriction of \(\sigma_h\) to the interior \(\Delta_r\) is a real analytic embedding and (iii) it is semi-algebraic. The complex structure automorphism \(J:T_{{}_P}X\to T_{{}_P}X\) acts on the tangent space at \(P \). For \(P\in \Delta \) the author looks at the real tangent space \(T_{{}_P}\Delta\) and to \(T_{{}_P}\Delta +J( T_{{}_P}\Delta )\), which is a complex subspace of \(T_{{}_P}X\), we have written here simply \( \Delta\) instead of \(\sigma_h(\Delta)\). For a positive integer \(m\) he considers the following condition on the complex dimension: \(\dim \bigl( T_{{}_P}\Delta +J( T_{{}_P}\Delta ) \bigr)\leq n-m \). He defines \(S_{m,r}\) to be the space of classes in \(H_{r}(X, \mathbb C)\) which are generated by singular \(r\)-simplexes as above, which moreover satisfy the said condition for all ponts \(P \in \sigma_h(\Delta)\). The author notes the inclusion \(N_{m ,r} \subset S_{m,r}\). The main theorem states \(N_{m ,r} = S_{m,r}\). The bulk of the proof is to show that the Zariski closure of a simplex as above is of dimension \(n-m\). The method rests on the study of the properties of semi-algebraic sets, the author declares that he drew inspiration for this from the work of \textit{E. Fortuna} and \textit{S. Łojasiewicz} [J. Reine Angew. Math. 329, 215--220 (1981; Zbl 0463.32009)].