Let $N$ be the total space of the vector bundle $\mathscr O_{\mathbb P^1}(-1)\oplus \mathscr O_{\mathbb P^1}(-1)$ over $\mathbb P^1$, and let $C_0\subset N$ be the zero section. Then $N$ is a quasi-projective Calabi-Yau threefold, and the moduli space of stable maps $\overline{\mathcal M}_g(N,[dC_0])$ is quasi-projective. It contains, as an open and closed substack, the space $$\overline{\mathcal M}_g(\mathbb P^1,d)\subset \overline{\mathcal M}_g(N,[dC_0]).$$ Faber and Pandharipande, in this paper on Hodge integrals, compute the contribution of the stable maps to $N$ factoring through the zero section $C_0\cong\mathbb P^1$ to the Gromov-Witten invariants of $N$. If the data $$\pi:U\to \overline{\mathcal M}_g(\mathbb P^1,d),\qquad f:U\to\mathbb P^1$$ describe the universal stable map to $\mathbb P^1$, this contribution is defined to be: $$C(g,d)=\int_{[\overline{\mathcal M}_g(\mathbb P^1,d)]^{\textrm{vir}}}e(R^1\pi_\ast f^\ast N).$$ In the paper, they write explicitely the generating series for these contributions in degree $1$. It is: $$\sum_{g\geq 0}C(g,1)t^{2g}=\Bigl(\frac{t/2}{\sin(t/2)} \Bigr)^2.$$

Question. Does anybody know (where to find) a similar formula for the generating series $\sum_{g\geq 0}C(g,d)t^{2g}$ when $d>1$? Maybe, it is hidden in that same paper, but I cannot see it. However, I am pretty sure this has been computed, as the Gromov-Witten theory of the local $\mathbb P^1$, as far as I know, is well understood.

Thank you very much!


1 Answer 1


The formula is given by

$$ \sum_{g\geq 0} c(g,d) t^{2g-2} = \frac{1}{d}\left( 2\sin \left(\frac{dt}{2}\right)\right)^{-2}$$

It is in Faber-Pandharipande http://arxiv.org/pdf/math/9810173v1.pdf. See also equations (1) and (2) in http://arxiv.org/pdf/math/9811140.pdf.

By the way, the inclusion you write $\overline{M}_g(\mathbb{P}^1,d) \subset \overline{M}_g(N,d[C_0])$ is actually an equivalence --- every stable map to $N$ lies in the zero section. The two moduli spaces are the same, but their virtual fundamental classes are different. The obstructions to deforming a map $f:C\to N$ off of the zero section lie in $H^1(C,f^*(N))$. These spaces are the fibers of a bundle whose Euler class capped against the virtual fundamental class of $\overline{M}_g(\mathbb{P}^1,d)$ gives the virtual class of $\overline{M}_g(N,d[C_0])$. Thus the integral you write for $C(g,d)$ is not a definition but actually a theorem (expressing the above relation of the virtual classes).

  • $\begingroup$ Thank you very much for your answer, and for adding the last clarifying paragraph! $\endgroup$
    – Brenin
    Commented Oct 4, 2014 at 17:34
  • $\begingroup$ I just wonder: how is, then, the contribution $C(g,d)$ actually defined? For instance: what about the "non-local" case of a rigid curve $C\subset Y$ of genus $g$ in a CY3 $Y$? It seems that the substack $\overline M_{g+h}(C,d[C])\subset \overline M_{g+h}(Y,d[C])$ being a connected component is essential for the contribution $C_g(h,d)$ to be well-defined. $\endgroup$
    – Brenin
    Commented Oct 4, 2014 at 17:42
  • $\begingroup$ Right, for the contribution of $C\subset Y$ to the GW invariants of degree $d[C]$, one needs the condition of $d$-rigidity. $O(-1)\oplus O(-1)$ curves are super-rigid, they satisfy the $d$-rigidity condition for all $d$. Precisely this issue is discussed in detail in one of my papers with Rahul: arxiv.org/pdf/math/0405204.pdf $\endgroup$
    – Jim Bryan
    Commented Oct 4, 2014 at 23:44

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.