Let $X$ be a smooth complex projective manifold, and $L$ an ample line bundle. By Bertini's Theorem, for every integer $q$ big enough there exists an open dense subset $U_q\subset |qL|$ such that every divisor $D$ in $U_q$ is smooth.

Warm up question: is the complement of $U_q$ always a divisor?

We can define a bigger open subset $V_q\subset |qL|$ as $$ V_q:=\{D\in |qL| \; \textrm{s. t.} \; (X,\frac{1}{q}D) \; \textrm{ is klt} \} $$

My question is: what is the dimension of the complement of $V_q$ ? (or at least can we bound its asymptotic in $q$ ? e.g. is it upper-bounded by $aq^{\dim X}$ with $a$ a constant which is strictly smaller than the volume of $L$? )

  • 3
    $\begingroup$ For all $q\geq q_0$, the complement of $U_q$ is a divisor whose degree is known, cf. mathoverflow.net/questions/165672/… Replace $\mathcal{O}(1)$ in that answer by an appropriately positive tensor power of $L$, and use the asymptotic formula to find $q_0$. $\endgroup$ Nov 20, 2018 at 14:41
  • $\begingroup$ Thanks! My main issue is with $V_q$. $\endgroup$
    – Giulio
    Nov 20, 2018 at 14:43
  • $\begingroup$ Just to make things more explicit, I am interested in the case $q>>0$. Moreover, let me explain a little computation to justify my last guess about the complement of $V_q$. Take $X=\mathbb{P}^2$, and $L=\mathcal{O}(3)$. Fix a line $H$, multiplying by $(q+1)H$ we can embed $| \mathcal{O}(2q-1)|$ into the complement of $V_q$. My guess is that this is kind of the biggest thing that can be outside $V_q$ (but I have no proof for the moment). $\endgroup$
    – Giulio
    Nov 20, 2018 at 20:37

1 Answer 1


Regarding the warm up question: No (although for $q$ sufficiently large the answer is yes, as Jason Starr comments.) Let $X = \mathbb{P}^1 \times \mathbb{P}^2$ and let $L= \mathcal{O}(1,1)$. Write homogeneous coordinates on the first factor as $(u:v)$ and on the second factor as $(x:y:z)$. A divisor $D$ in $H^0(L)$ is of the form $$u (ax+by+cz) + v (dx+ey+fz)=0.$$ $D$ is singular if and only if the matrix $$\begin{bmatrix} a & b& c \\ d & e & f \\ \end{bmatrix}$$ has rank $1$. (If this matrix has rank $2$, $D$ is isomorphic to $\mathbb{P}^2$ blown up at a point, when the matrix drops rank, $D$ turns into the union of a $\mathbb{P}^2$ and a $\mathbb{P}^1 \times \mathbb{P}^1$.) The condition that this matrix drops rank is codimension $2$.

The complement of $U_q$ is called the dual variety to $X$. It is usually (no precise meaning attached) true that the dual of a smooth variety is a divisor. For a while, it was an open problem to characterize smooth projective toric varieties whose duals were not divisors; this was solved by Dickenstein, Feichtner and Sturmfels.

  • $\begingroup$ Thanks!! In general, I am more interested in the asymptotic values of $q$ (actually, for $q=1$, $L$ might even have no sections at all!!). And of course, I think $V_q$ is the hard part. Thanks!! $\endgroup$
    – Giulio
    Nov 20, 2018 at 20:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.