math.AG — Algebraic Geometry
Contributed by David Jordan
Christol conjectured that every globally bounded, D-finite power series with rational coefficients is the diagonal of a rational function. We prove this for the hypergeometric series 3F2(a1,a2,a3;b,1;x) with rational parameters such that ai, b and ai−b are not integers: if such a series is globally bounded, then it is the diagonal of an algebraic power series in two variables, and hence of a rational function in three variables; three is the least possible number of variables. This includes Christol's example 3F2(91,94,95;31,1;x) from 1987, and the series 3F2(71,72,74;21,1;x), which Christol suggested as a possible counterexample to his conjecture. By a result of Rivoal and Roques, under the Rohrlich-Lang conjecture the latter is a diagonal that is not a Hadamard product of algebraic power series. The proof is Hodge-theoretic and not constructive: it shows that such diagonal representations exist, but does not produce them. Using Fedorov's formula for the Hodge numbers of hypergeometric local systems, we show that the series is globally bounded exactly when every Galois conjugate of its local system carries a complex variation of Hodge structure with two adjacent Hodge levels. The restriction of scalars of the local system then underlies a polarized integral variation of Hodge structure of weight one. By a theorem of Deligne the series is therefore a period of an abelian scheme, and hence of a family of curves, where it is an integral over vanishing cycles of a semistable degeneration. Expanding this integral at the nodes of a regular model gives the diagonal representation.
Primarily AI-generated textHuman understanding: no partsChristol’s conjectureabelian schemesdiagonals of rational functionsglobally bounded power serieshypergeometric functionsvanishing cyclesvariations of Hodge structure