A correction: ``On a conjecture of Clemens on rational curves on hypersurfaces'' (Q5960512)

In the previous paper cited in the title [\textit{C. Voisin}, J. Differ. Geom. 44, No. 1, 200-213 (1996; Zbl 0883.14022)], the author stated the following result, which implies a conjecture by H. Clemens and generalizes a previous result by L. Ein: Let \(X \subset \mathbb{P}^n\) be a general hypersurface of degree \(d\geq 2n-k-1\), with \(1\leq k\leq n-3\); then any subvariety \(Y\) of \(X\) of dimension \(k\) has a desingularization \(\widetilde{Y}\) with an effective canonical bundle; if the inequality is strict, the sections of \(K_{\widetilde{Y}}\) separate generic points of \(Y\). Let \({\mathcal X}\subset \mathbb{P}^n\times S_d\), \(S_d=H^0({\mathcal O}_{{\mathbf P}^n}(d))\), be the universal hypersurface, then Ein's result follows from the fact that \(\wedge^{n-k-1}T_{\mathcal X}(n-k-1)|_{X_F}\) is generated by global sections (on the smooth fiber \(X_F\), \(F\in S_d\)). The author's previous proof used the hypothesis that, for generic \(F\), the space \(H^0(\wedge ^2 T_{\mathcal X}(1)|_{X_F})\) has no base points on the set of subspaces \(V\subset T_{{\mathcal X},(x,F)}\) containing the tangent space to the \(Gl(n+1)\)-orbit of \((x,F)\). However, this hypothesis is not true (as examples show). The present paper gives a new proof of the result, taking into account the spaces \(V\) in the base locus of \(H^0(\wedge ^2 T_{\mathcal X}(1)|_{X_F})\) and describing them.