Rational curves of minimal degree and characterizations of projective spaces (Q2498482)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Rational curves of minimal degree and characterizations of projective spaces |
scientific article |
Statements
Rational curves of minimal degree and characterizations of projective spaces (English)
0 references
16 August 2006
0 references
Let \(X\) be a smooth complex projective variety; \(X\) is said to be uniruled if there exists a rational curve through every point of \(X\). If \(X\) is uniruled there exist irreducible components of the scheme RatCurves\(^n(X)\) such that the corresponding universal families dominate \(X\); such components are called dominating families of rational curves. A dominating family \(H\) is called minimal if for a general \(x \in X\) the subfamily \(H_x\) parametrizing curves of \(H\) through \(x\) is proper. If \(X\) is uniruled then on \(X\) there exists a minimal dominating family of rational curves (for instance one can take a dominating family of minimal degree with respect to some fixed ample line bundle). \textit{J.-M. Hwang} and \textit{N. Mok} recently [in: School on vanishing theorems and effective results in algebraic geometry. Lect. notes school Trieste, Italy, 2000. ICTP Lect. Notes 6, 335--393 (2001; Zbl 1086.14506)] started the study of the subvariety of tangent directions \({\mathcal C}_x \subset \mathbb P (T_xX)\), which is defined as the closure of the image of the tangent map \(\tau_x:H_x --> \mathbb P (T_xX)\), defined by sending a curve which is smooth at \(x\) to its tangent direction at \(x\); by results of \textit{S. Kebekus} [J. Algebr. Geom. 11, 245--256 (2002; Zbl 1054.14035)] and \textit{J.-M. Hwang} and \textit{N. Mok} [Asian J. Math. 8, 51--64 (2004; Zbl 1072.14015)] the map \(\tau_x:H_x \to {\mathcal C}_x\) is the normalization. The paper under review studies varieties for which, for a general \(x \in X\), \({\mathcal C}_x\) is a union of linear \(d\)-dimensional subspaces of \(\mathbb P(T_xX)\), showing that in this case there exists a variety \(X'\), with a finite morphism onto \(X\), a dense open subset \(U^0\) of \(X'\) and a \(\mathbb P^{d+1}\)-bundle \(\varphi^0:U^0 \to T^0\) (for a more precise statement see Theorem 3.1 and the discussion above). In case \({\mathcal C}_x\) is a linear \(d\)-dimensional subspace of \(\mathbb P(T_xX)\) for a general \(x \in X\), then there exists a dense open subset \(U^0\) of \(X\) and a \(\mathbb P^{d+1}\)-bundle \(\varphi^0:U^0 \to T^0\) such that any curve of \(H\) meeting \(X^0\) is a line in a fiber of \(\varphi^0\) (Theorem 1.1). A first application of this result is a unified geometric proof of the characterization of the projective space as the only smooth complex projective variety \(X\) whose tangent bundle \(TX\) contains an ample locally free subsheaf \(E\). The case \(E \simeq TX\) is a celebrated result of \textit{S. Mori} [Ann. Math. (2) 110, 593--606 (1979; Zbl 0423.14006)]; the case in which \(E\) is a line bundle was considered by \textit{J. M. Wahl} [Invent. Math. 72, 315--322 (1983; Zbl 0544.14013)] with completely different methods. \textit{F. Campana} and \textit{T. Peternell} [Manuscr. Math. 97, No.~1, 59--74 (1998; Zbl 0932.14024)] settled the cases in which rank \(E \geq \dim X-2\), and the proof was completed for every rank by \textit{M. Andreatta} and \textit{J. A. Wiśniewski} [Invent. Math. 146, 209--217 (2001; Zbl 1081.14060)]; this last proof relies on Mori's and Wahl's results. A second application of Theorem 1.1 is a characterization of products of projective spaces: Let \(X\) be a smooth complex projective variety with \(k\) distinct unsplit dominating families of rational curves \(H_1, \dots, H_k\) such that, for a general point \(x \in X\) the subvarieties of tangent directions at \(x\) to curves of the families are linear subspaces of dimension \(d_i\) with \(\sum d_i = \dim X -k\). Then \(X \simeq \mathbb P^{d_1 +1} \times \dots \times \mathbb P^{d_k +1}\). To prove this the author shows that, in the above assumptions, the numerical classes of the families \(H_1, \dots, H_k\) are linearly independent in \(N_1(X)\) and thus she can apply a result of the reviewer [Can. Math. Bull. 49, 270--280 (2006; Zbl 1115.14034)].
0 references