Let $\overline{M}_{0,k}(\mathbb{P}^n,d)$

denote the moduli space of genus zero degree $d$ stable maps with $k$ marked points. This is an orbifold of expected dimension. Let $\overline{U}_{0,k}(\mathbb{P}^n,d)$ be the corresponding universal curve. Then we have two morphisms

$$\pi_1: \overline{U}_{0,k}(\mathbb{P}^n,d) \rightarrow \overline{M}_{0,k}(\mathbb{P}^n,d)$$


$$\pi_2: \overline{U}_{0,k}(\mathbb{P}^n,d) \rightarrow \mathbb{P}^n$$

So if we have a coherent sheaf $F$ on $\mathbb{P}^n$ we can pull it back to universal curve and push it forward to moduli space, i.e. we can consider $F_i=(R^i\pi_{1*})(\pi_2^*F)$ on moduli space.

Here are my questions:

1) Are $F_i$ zero for $i>0$ ? (is $\pi_{1*}$ exact )

2) If $F$ is equal to ideal sheaf of some smooth projective subvariety $X$, and $\beta \in H_2(X)$ is of degree $d$, then is there in general any relation between some component of support of $F_0$ and the moduli space $\overline{M}_{0,k}(X,\beta)$ ?

3) if the answer to question one is No, what is the interpretation of $F_i$, $i>0$, for the case when $F$ is the ideal sheaf of $X$ as in question two?

  • $\begingroup$ I am not sure if on your PC, you can see the Latex part correctly. Whatever I do, it does not show the line defining $\pi_1$ correctly. $\pi_1$ is the obvious map from universal curve to moduli space. $\endgroup$ Jun 23, 2011 at 16:20
  • $\begingroup$ Motivation: This technique was used by some people to construct a virtual cycle for moduli spaces corresponding to quintic in $\mathbb{P}^4$ and using that plus localization they computed the GW invariants of quintic. There $F_0$ is a vector bundle on $\overline{M}(\mathbb{P}^4,d)$ and the zero locus of some section of that is a virtual fundamental cycle for degree $d$ maps to quintic. $\endgroup$ Jun 23, 2011 at 16:55

1 Answer 1


1) It depends on $F$. Roughly speaking fibers of $\pi_1$ are the curves parametrized by $\overline{M}_{0,k}(P^n,d)$. So if you want $R^{>0}\pi_{1*}\pi_2^*F$ to vanish, you have to check that $H^{>0}$ of the restriction of $F$ to any such curve vanishes. Since all the curves are rational it suffices (but not necessary) to assume that $F$ is generated by global section.

2,3) It seems that you want something else. The usual approach is the following. Assume that $X$ is the zero locus of a regular section $s$ of a vector bundle $F$. Then $s$ gives a global section of $\pi_2^*F$, and hence of $\pi_{1*}\pi_2^*F$. The claim is that (the virtual fundamental class of) $\overline{M}_{0,k}(X,\beta)$ is the zero locus of this section of $\pi_{1*}\pi_2^*F$ on $\overline{M}_{0,k}(P^n,d)$.

  • $\begingroup$ You did not answer my 3rd question. What is the geometric meaning of higher derived sheaves? If we start with a vector bundle whose zero section is $X$ then what you say is true, but I'd like to start from the ideal sheaf of $X$ itself and then look at the support of induced sheaf on moduli space. In fact I want to know if this gives the same picture as in the case you mentioned. $\endgroup$ Jun 23, 2011 at 19:03
  • $\begingroup$ Does anyone know a good reference for the fact in 2) that the virtual fundamental class of $\overline{M}_{0,k}(X,\beta)$ is the zero locus of that section? $\endgroup$ Sep 21, 2012 at 8:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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