Lefschetz classes of simple factors of the Jacobian variety of a Fermat curve of prime degree over finite fields (Q1938560)

Let \(A\) be an abelian variety defined over a finite field, \(\overline{A}\) be its base change to an algebraic closure of the field of definition. A Tate class of \(A\) is by definition a class in \(H^{2i}_{\mathrm{\'et}}(\overline{A},\mathbb{Q}_{\ell}(i))\) which is Galois-invariant. An algebraic class is a class in the image of the class cycle map (i.e. it can be written as a \(\mathbb{Q}_{\ell}\)-linear combination of classes of subvarieties of \(A\) of codimension \(i\)). Clearly an algebraic class is a Tate class and the Tate conjecture predicts that the converse is true (the conjecture more generally applies to smooth and projective varieties defined over a field which is finitely generated over its prime field, for instance finite fields or number fields). This conjecture is still open. The case \(i=1\) is known by the work of \textit{J. Tate} [Invent. Math. 2, 134--144 (1966; Zbl 0147.20303)]. Hence, one defines Lefschetz classes in \(H^{2i}_{\text{ét}}(\overline{A},\mathbb{Q}_{\ell}(i))\) as those spanned by cup products of \(i\) Tate classes in \(H^{2}_{\text{ét}}(\overline{A},\mathbb{Q}_{\ell}(1))\) and deduces that all Lefschetz classes are algebraic. In particular, Tate conjecture for \(A\) is true if all Tate classes are Lefschetz classes. This equality is easier to be checked as it does not deal anymore with algebraic cycles but just with cohomological classes (on the other hand it is not verified for all abelian varieties, one should think of it as a lucky situation). In the paper under review the author gives some conditions on \(A\) (Theorem 3.1 and Theorem 3.4) that imply the equality between Tate classes and Lefschetz classes (and in particular Tate conjecture for \(A\)). This allows him to give some explicit cases where this equality holds (Theorem 1.1). For example the equality holds when \(A\) is an absolute simple factor of the Jacobian of the Fermat curve of degree \(m\), assuming that \(m\) is an odd prime different from the characteristic of the field of definition and that the degree over \(\mathbb Q\) of the center of the endomorphism algebra \(\mathrm{End}(\overline{A})\otimes \mathbb{Q}\) is not divisible by \(6\). The main ingredient in the proofs is the following result of \textit{J. S. Milne} [Duke Math. J. 96, No. 3, 639--675 (1999; Zbl 0976.14009)]. Consider the Lefschetz group \(L(A)\), algebraic subgroup of \(\mathrm{GL} (H^{1}_{\text{ét}}(\overline{A},\mathbb{Q}_{\ell}))\) defined as the group of automorphisms of \(H^{1}_{\text{ét}}(\overline{A},\mathbb{Q}_{\ell})\) commuting with \(\mathrm{End}(\overline{A})\otimes \mathbb{Q}_{\ell}\) and respecting up to a scalar the pairing induced by a polarization (the definition does not depend on the choice of the polarization). Let \(P(A) \subseteq L(A)\) be the Zariski closure of the image of the Galois group. Then all Tate classes are Lefschetz classes if and only if \(P(A) =L(A).\) This kind of questions were studied in the Hodge setting by several authors. An example is the following theorem of \textit{S. G. Tankeev} [Math. USSR, Izv. 20, 157--171 (1983); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 46, No.1, 155-170 (1982; Zbl 0587.14005)], generalized by \textit{K. A. Ribet} [Am. J. Math. 105, 523--538 (1983; Zbl 0586.14003)]: if \(A\) is a complex abelian variety of prime dimension, whose endomorphism algebra \(\mathrm{End}(\overline{A})\otimes \mathbb{Q}\) is a quadratic imaginary field, then all Hodge classes of \(A\) are Lefschetz classes, and in particular Hodge conjecture holds for \(A\).