We consider the Zariski-Lipman Conjecture on a free module of derivations for algebraic surfaces. Using the theory of non-complete algebraic surfaces, and some basic results about ruled surfaces, we will prove the conjecture for several classes of affine and projective surfaces.THE ZARISKI-LIPMAN CONJECTURE
271As far as we can see, none of these results imply our results. Given a smooth algebraic surface V and a projective completionV of it such that D :=V − V is a divisor whose only singularities are nodes, the logarithmic Kodaira dimension of V , denoted byκ(V ), is defined as the supremum of the dimensions of the images ofV under the rational maps defined by H 0 (V , n(K + D)), n 1. If this linear system is trivial for all n 1, then we defineκ(V ) = −∞ (see [11]).Our approach to this conjecture in this paper is global and it uses the theory of non-complete algebraic surfaces developed by Iitaka, Kawamata, Fujita, Miyanishi, Sugie, Tsunoda and other Japanese mathematicians. This theory has proved to be very effective in the solution of many problems about non-complete algebraic surfaces, and the arguments in this paper are just one more instance of this. We also use some results from the theory of vector bundles on smooth projective curves. Although some of our arguments are valid assuming only projectivity of the module of derivations, for many arguments we need the full force of the assumption that the tangent bundle of the smooth locus is trivial. As can be seen from the somewhat involved proofs in this paper, this stronger hypothesis is justified by the difficulty of the general conjecture. In this paper we verify the conjecture for all affine surfaces V such thatκ(V − Sing(V )) 1, and prove the conjecture in almost all the cases when V is projective. In particular, we prove that if V is projective, the tangent bundle of V − Sing(V ) is trivial andκ(V − Sing(V )) = 0 or 1, then V is smooth.All the varieties in this paper will be assumed to be over an algebraically closed field k of characteristic 0. If V is an algebraic surface, then by V 0 we denote the smooth locus of V .We now state the results of this paper.Theorem 1.2. Let V be an algebraic surface defined over k. Assume that the tangent bundle of V 0 is trivial. The following statements hold.(1) If V is an affine algebraic surface such thatκ(V 0 ) 1, then V is smooth.(2) If V is a projective surface, thenκ(V 0 ) 0 and V has at most one singularity.(3) If V is a projective surface such thatκ(V 0 ) = 0, then V is smooth.(4) Assume that V is a projective surface such thatκ(V 0 ) = −∞. Let p be the unique singular point of V . Then there exists a resolution of singularities π : W → V such that there is a P 1 -fibration W → C, where C is a smooth projective curve. If the genus of C is at least 2, then V is smooth.(5) With the notation in (4), if C is a rational curve, then V is smooth.(6) With the notation in (4), let C be an elliptic curve. If at least one singular fiber of W → C has a non-reduced feather (defined later), then V is smooth. Remark 1.3. We ...