3
$\begingroup$

Let us work over an algebraically closed field $K$. If $X\subset \mathbb{P}^n$ is a closed subset of dimension $r$, then there should exist a linear subspace $L\subset \mathbb{P}^n$ of dimension $n-r$ such that $X\cap L=\{x_1,\ldots,x_k\}$ is finite, and such that for each $i\in \{1,\ldots,k\}$ the tangent spaces of $X$ and $L$ at $x_i$ intersect only at $0$ (or at $x_i$ if you look at the projective tangent spaces, embedded in $\mathbb{P}^n$).

Is there an easy proof of this existence?

Over $K=\mathbb{C}$, in Mumford, Algebraic Geometry, Complex Algebraic Geometry, Theorem 5.1, not only the existence is claimed, but also the fact that the number $k$ is constant (equal to the degree of $X$, either as a definition, or as a theorem, depending on how you define the degree). The fact that $k$ is constant is explained in a long proof, and the existence is proven by taking a linear subspace $L'\subset \mathbb{P}^n$ of dimension $n-r-1$, projecting from $L'$ and taking a point of $\mathbb{P}^r$ over which the projection is smooth. However, the fact that the projection $X\dashrightarrow \mathbb{P}^r$ is smooth on a general point of $\mathbb{P}^r$ is claimed without any argument. Why is it true? It seems to me that it is false in positive characteristic, but maybe it is true for a right choice of $L'$ ? Or using another proof?

$\endgroup$
7
  • 1
    $\begingroup$ Is it not essentially purity of the branch locus (at least when $X$ is smooth)? $\endgroup$ Commented Feb 27, 2021 at 19:40
  • $\begingroup$ Do you mean for the morphism $X\to \mathbb{P}^r$ ? If $X$ is the smooth conic $x^2+yz=0$ in $\mathbb{P}^2$ over an algebraically closed field of characteristic $2$, and $p$ is the point $[1:0:0]$ then the projection from $p$ gives a morphism $X\to \mathbb{P}^1$ which is not smooth (every fibre consists of one point of multiplicity $2$), so in positive caracteristic the statement is false if you take "any" L'. Why is it true for "some" L' ? By the way, I was not necessary asked $X$ to be smooth (and neither is Mumford, but he works over $\mathbb{C}$). $\endgroup$ Commented Feb 27, 2021 at 19:47
  • $\begingroup$ Maybe I am missing something, but Mumford's book only considers the complex case, hence characteristic $0$. In this situation, purity of the branch locus prevents examples as the one of the conic. $\endgroup$ Commented Feb 27, 2021 at 19:48
  • 1
    $\begingroup$ Can one do induction on dimension, using the following: intersecting $X$ with a general hyperplane through a nonsingular point $P \in X$ is also nonsingular at $P$? The base of induction would be that a generically reduced plane curve has a line which is nowhere tangent. $\endgroup$ Commented Feb 28, 2021 at 0:37
  • 2
    $\begingroup$ This is just a dimension count: it is more or less obvious that a general linear subspace will intersect $X$ along its smooth locus (the complement has smaller dimension) and for a smooth point the dimension of the set of linear spaces not intersecting transversally at that point is independent of the point. $\endgroup$
    – naf
    Commented Feb 28, 2021 at 7:41

0

You must log in to answer this question.