Periods of Hodge cycles and special values of the Gauss' hypergeometric function (Q2162791)

In this paper, the author studies the Hodge theory of certain perturbations of Fermat varieties, computing the periods of certain \textit{strong generic Hodge cycles} (Definition 4). These are a certain explicitly constructed subset of the \textit{generic Hodge cycles}, which are the Hodge classes that remain of Hodge type under parallel transport along any smooth perturbation by a polynomial in one variabie. By an observation of Deligne, if \(X\) is a smooth projective variety defined over a subfield \(k\subset \mathbb{C}\), then the periods of any algebraic cycles with respect to algebraic differential forms defined over \(k\) are always algebraic over \(k\), up to a certain multiple of \(2\pi i\). Deligne also proved that the same holds true for Hodge cycles in classical Fermat varieties defined by polynomials of the form \(x_1^d+\cdots+x_d^{n+1}\), even though the Hodge conjecture is unknown in that case. The author extends this idea to strong generic Hodge cycles on certain varieties defined by polynomials \(x_1^{m_1}+\cdots x_n^{m_n}+y(y-1)(y-\lambda)\), proving similar algebraicity results. By explicitly computing examples, it is shown that the hypergeometric functions \(F(\frac{5}{6},\frac{1}{6},\frac{5}{3};1-\lambda)\) and \(F(\frac{7}{6},-\frac{1}{6},\frac{7}{3};1-\lambda)\) are algebraic over \(\mathbb{Q}(\lambda)\). The author also finds a number of non-trivial algebraic relations between hypergeometric functions, where the functions themselves are not algebraic over \(\mathbb{Q}(\lambda)\) (see Proposition 1). The algebraicity result is obtained by reducing computations of \(n\)-dimensional integrals over strong generic Hodge cycles, to \(2\)-dimensional integrals over \((1,1)\)-classes, which are algebraic by the Lefschetz \((1,1)\)-classes.