Pole structure of topological string free energy (Q2495267)

The paper provides a combinatorial proof of the Gopakumar-Vafa conjecture for the total space of the canonical bundle on a smooth complete toric surface \(S\). More precisely, if \(X\) is a Calabi-Yau threefold then according to the Gopakumar-Vafa conjecture, the generating function of the Gromov-Witten invariants of \(X\) can be written as \[ {\mathcal F}_\beta(X)=\sum_{g\geq 0}\sum_{k| \beta}\frac{n^g_{\beta/k}(X)}{k}\left(2\sin \frac{k g_s}{2}\right)^{2g-2}, \] where the \(n^g_\beta(X)\) are integers (which equal the number of certain BPS states in M-theory) and vanish for fixed \(\beta\) and \(g\gg 0\). Via the Möbius inversion formula, this identity can be rewritten as \[ \sum_{k| \beta}\frac{\mu(k)}{k}{\mathcal F}_{\beta/k}(X)\biggr|_{g_s\to kg_s}=\sum_{g\geq 1}n^g_\beta(X)\left(2\sin \frac{g_s}{2}\right)^{2g-2} \] Let \(t=(q^{1/2}-q^{-1/2})^2\). Then the elementary identity \( \left(2\sin \frac{g_s}{2}\right)^2= -t\,\bigr|_{q=e^{\sqrt{-1} g_s}} \) implies that (forgetting its M-theoretical aspects) the Gopakumar-Vafa conjecture is equivalent to saying that \[ \sum_{k| \beta}\frac{\mu(k)}{k}{\mathcal F}_{\beta/k}(X)\biggr|_{g_s\to kg_s}=t^{-1}p(t)\,\bigr|_{q=e^{\sqrt{-1} g_s}} \] with \(p(t)\in{\mathbb Z}[t]\). In case \(X\) is the total space of the canonical bundle on a smooth complete toric surface \(S\), one can compute Gromov-Witten invariants via localization, obtaining a combinatorial expression for the generating function: \[ {\mathcal F}_\beta(X)=\sum_{[\vec{d}\cdot \vec{C}]=\beta}{\mathcal F}^{\gamma_S}_{\vec{d}}(q)\, \bigr|_{q=e^{\sqrt{-1}g_s}}, \] where \(\vec{C}=(C_1,\dots,C_r)\) are the toric invariant rational curves of \(S\), \(\gamma_S=({C_1}^2,\dots,{C_r}^2)\) is the vector of self-intersection numbers and \({\mathcal F}^{\gamma}_{\vec{d}}(q)\) is a combinatorial expression written in terms of skew-Schur functions [\textit{C.-C.~M.~Liu, K.~Liu, J.~Zhou}, \texttt{math.AG/0310272} and \textit{J.~Zhou}, \texttt{math.AG/0310283}]. Defining the combinatorial function \[ G_{\vec d}^\gamma(q)=\sum_{k'| k}\frac{k'}{k}\mu\left( \frac{k}{k'}\right){\mathcal F}^\gamma_{k'\vec{d}/k}(q^{k/k'}),\qquad k=\text{ gcd}(\vec{d}), \] the Gopakumar-Vafa conjecture for \(X=K_S\) becomes \[ \sum_{[\vec{d}\cdot\vec{C}]=\beta}G^{\gamma_S}_{\vec{d}}(q)\, \bigr|_{q=e^{\sqrt{-1}g_s}}=t^{-1}p(t)\,\bigr|_{q=e^{\sqrt{-1} g_s}} \] and so it is implied by the purely combintorial statement \(t\, G^\gamma_{\vec{d}}(q)\in {\mathbb Z}[t]\). Actually, since \textit{P. Peng} has shown [\texttt{math.AG/0410540}] that \(G^\gamma_{\vec{d}}(q)\) is a rational function in \(t\) such that its numerator and denominator are polynomials with integer coefficients, with a monic denominator, one only needs to prove that \(t\, G^\gamma_{\vec{d}}(q)\in {\mathbb Q}[t]\), which is the main result in the paper. The proof goes as follows. The functions \({\mathcal F}^{\gamma}{\vec{d}}(q)\) are the coefficients of the topological string free energy \({\mathcal F}^\gamma(\vec{Q})=\sum_{\vec{d}}{\mathcal F}^\gamma_{\vec{d}}(q) \vec{Q}^{\vec{d}}\). Using Okounkov's formalism of infinite-wedge spaces and fermion operator algebras, one can write the partition function \({\mathcal Z}^\gamma(\vec{Q})=\exp{\mathcal F}^\gamma(\vec{Q})\) in terms of matrix elements of a certain operator [\textit{A.~Okounkov}, Sel. Math., New Ser. 7, No.1, 57--81 (2001; Zbl 0986.05102) and \textit{A.~Okounkov} and \textit{R.~Pandharipande}, Ann. Math. (2) 163, No. 2, 561--605 (2006; Zbl 1105.14077)]. These matrix elements can be expressed as amplitudes of suitable Feynman diagrams, so that the partition function is written as a sum over graphs and the free energy as a sum over connected graphs. Consequently, the functions \(G^\gamma_{\vec{d}}(q)\) are expressed as sums over connected graphs. The contribution of a graph with strictly positive first Betti number to \(G^\gamma_{\vec{d}}(q)\) is a polynomial in \(t\) with rational coefficients. The contribution from a tree, instead, is an element of \({\mathbb Q}(t)\) with non-trivial poles. Yet, a careful analysis of the pole structure shows that in the sum all pole contributions cancel except possibly a simple pole at \(t=0\).