Rationally connected varieties over finite fields (Q1423873)

The main object of the paper under review is a separably rationally connected variety \(X\) defined over a field \(k\) (usually assumed finite or local), that is a smooth, proper variety admitting a morphism \(U\times {\mathbb P}^1\to X\) such that the induced map \[ F(-,(0:1))\times F(-,(1:0)) \colon U\to X\times X \] is dominant and separable (roughly, this means that one can join a given pair of points by a rational curve). The authors' goal is to prove that if \(k\) is a finite field, then such an \(X\) contains ``many'' rational curves. More precisely, they prove that if the dimension of \(X\) is at least 3, \(S\subset X\) is a zero-dimensional smooth subscheme, and the cardinality of \(k\) is bigger than some number depending on the dimension of \(X\) and the degrees of \(X\) and \(S\), then there exists a smooth rational curve \(C_S\subset X\) containing \(S\). This result has some striking applications to the case when \(k\) is a number field (thus answering an old question of J.-L.~Colliot-Thélène): it implies that for almost all primes \(v\) of \(k\) the R-equivalence on \(X(k_v)\) is trivial and the group of classes of zero-cycles of degree zero of \(X_v\) equals zero. This has been known only for some special classes of varieties like rational surfaces or linear algebraic groups. (For some more recent development, see \textit{J.~Kollár} [Publ. Res. Inst. Math. Sci. 40, 689--708 (2004; Zbl 1081.14008)] and \textit{J.-L.~Colliot-Thélène} [Invent. Math. 159, 589--606 (2005; Zbl 1080.14012)]). The method of the proof roughly looks as follows. The aim is to find a rational curve on \(X\) passing through a given \(S\). There are such curves defined over \(\bar k\), hence the space \(M\subset \text{Hilb}(X)\) of smooth rational curves containing \(S\) is not empty. To prove that \(M(k)\) is not empty, the authors first find an irreducible subvariety \(W\subset M_{\bar k}\) which is Galois invariant. This is possible if \(k\) is an arbitrary perfect field: the proof uses a geometric method of ``combs'' developed by the first author and is the most technical part of the paper. To deduce the existence of a rational point on \(M\), the authors suggest two ways: either to use the fact that the ultraproduct of infinitely many finite fields whose orders go to infinity is a pseudo-algebraically closed field (and hence every geometrically irreducible variety has a dense set of rational points), or to produce a bound on the basic projective invariants of \(W\) and use the Lang--Weil theorem.