MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Suppose we have a full exceptional collection (F1,...,Fn) of coherent sheaves on a smooth projective variety X. The number n of sheaves in this collection is equal to the rank of the Grothendieck group K0(X).

Is there any relation between n and the rank of the Picard group Pic(X) or the dimension of X?

share|cite|improve this question
If $X = \mathbb{P}^n,$ then its Picard group has rank one for any n, but a full exceptional sequence consists of n+1 line bundles. So I doubt there is any relation in general between the two. – Mike Skirvin May 26 '10 at 13:00
One simple relation is $n \ge \dim X + 2\rho(X) - 1$ (if $\dim X \ge 3$), where $\rho(X)$ is the rank of the Picard group. – Sasha May 26 '10 at 19:19
up vote 10 down vote accepted

As a far as I know, there is no obvious relation between the number of objects in a full exceptional collection and the dimension of $X$, unless you impose extra conditions on the full exceptional collection, as in Bondal and Polishchuk's 'Homological properties of associative algebras', where they prove that when the collection is 'geometric', then ${\rm rank}(K_{0})(X)={\rm dim} X + 1$. Their hypotheses apply for example to $\mathbb{P}^{n}$, where the standard full exceptional collection is $\mathcal{O},\mathcal{O}(1), ..., \mathcal{O}(n)$.

But for examples like $\mathbb{P}^{1} \times \mathbb{P}^{1}$ these hypotheses do not apply, since the standard full exceptional collection is $\mathcal{O}, \mathcal{O}(1,0), \mathcal{O}(0,1), \mathcal{O}(1,1)$, so ${\rm rank}(K_{0}(X))={\rm dim} X + 2$.

In general, a variety $X$ will not admit a full exceptional collection of objects in its derived category of coherent sheaves, for instance because $K_{0}(X)$ is usually not free of finite rank. And some varieties cannot carry any exceptional objects at all. For instance, if $X$ is smooth projective with trivial canonical bundle, then if you checked that $Hom(E,E)$ were one dimensional as needed for $E$ to be exceptional, Serre duality would imply that $Ext^{n}(E,E)$ is also one dimensional.

To me it seems more natural to ask for a nice 'compact generator' of the derived category of a variety. This is a perfect complex $E$ on $X$ (one locally quasi-isomorphic to a finite length complex of vector bundles) that generates the category $Perf(X)$ of perfect complexes in finitely many steps for each object or $D_{QCoh}(X)$, the unbounded derived category of quasi-coherent sheaves, in infinitely many steps. This is a generalization of the sum $E$ of the objects in a full exceptional collection, with no extra conditions imposed on the $Ext$ algebra of $E$.

Such compact generators exist in great generality, due to work of Bondal and van den Bergh. Given such a compact generator and letting $A=RHom(E,E)$, the derived endomorphism dg algebra of $E$, we get equivalences $RHom(E,?): D_{QCoh}(X) \rightarrow D(A)$ and $RHom(E,?): Perf(X) \rightarrow Perf(A)$, just as we would if $E$ were the sum of objects in a full exceptional collection.

One could ask if you can find concrete examples of such compact generators, and indeed you can. If $X$ is projective of dimension $d$, let $L$ be an ample, globally generated line bundle. Then it is easy to show using a standard criterion for compact generation that $E=\mathcal{O}\oplus L \oplus \cdots \oplus L^{d}$ is a compact generator of the derived category of $X$.

So one can always say that if $X$ is projective, then the minimal number of line bundles (not necessarily exceptional) needed to generate the derived category of $X$ is bounded by ${\rm dim} X + 1$.

I would also guess that fewer line bundles will not suffice, but I'm not sure.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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