Primitive forms for affine cusp polynomials (Q2286715): Difference between revisions

A primitive form is a generalization of a differential of the first kind on an elliptic curve. Their existence for isolated hypersurface singularities was proved in [\textit{M. Saito}, Ann. Inst. Fourier 39, No. 1, 27--72 (1989; Zbl 0644.32005)]. However, the explicit form of a primitive form is rarely known except for a few cases. This paper adds a new class of polynomials starting from a cusp polynomial $f_A$ of type $A$: \[ f_A(\mathbf{x})=x_1^{a_1}+x_2^{a_2}+x_3^{a_3}-q^{-1}x_1x_2x_3 \] where $A=(a_1,a_2,a_3)$ is a set of positive integers with $a_1\le a_2\le a_3$. Set $\mu_A=a_1+a_2+a_3-1$ and $\chi_A=1/a_1+1/a_2+1/a_3-1$, $f_A$ is called an affine cusp polynomial if $\chi_A>0$, which is treated in this paper. Then a primitive form for a universal unfolding $F_A$ of $f_A$ is determined. $F_A$ is a deformation of $f_A$ over a $\mu_A$-dimensional manifold $M=\mathbb{C}^{\mu_a-1}\times(\mathbb{C}\setminus\{0\})$, whose coefficient of the term $x_1x_2x_3$ is $-s_{\mu_a}^{-1}$: \[ F_A(\mathbf{x};\mathbf{s},s_{\mu_A})=x_1^{a_1}+x_2^{a_2}+x_3^{a_3}-s_{\mu_A}^{-1}\cdot x_1x_2x_3+s_1\cdot 1+\sum_{i=1}^3\sum_{j=1}^{a_1-1}s_{i,j}\cdot x_i^j \] Let $p_\ast\mathcal{O}_C$ be \[ \mathcal{O}_M[x_1,x_2,x_3]/\bigl(\frac{\partial F_A}{\partial x_1},\frac{\partial F_A}{\partial x_2}, \frac{\partial F_A}{\partial x_3}\bigr). \] Then it is the direct image of the sheaf of relative algebraic functions on the relative critical set $C$ of $F_A$ with respect to the projection $p:\mathbb{C}^3\times M\to M$. Then $F_A(\mathbf{x};\mathbf{0},q)=f_A(\mathbf{x})$ and the Kodaira-Spencer map $\rho:\mathcal{T}_M\to p_\ast\mathcal{O}_C$ is an isomorphism (Proposition 2.5). Based on this characterization, the filtered de Rham cohomology group $\mathcal{H}_{F_A}$ of the universal unfolding $F_A$ (Definition 2.17), Gauss-Manin connection $\nabla$ (Definition 2.27) and higher residue pairing $K_{F_A}$ (Definition 2.34) are defined. Then the main theorem of this paper is \textbf{Theorem 3.1}. The element $\zeta_A=[s_{\mu_a}^{-1}dx_1\wedge dx_2\wedge dx_3]\in \mathcal{H}^{(0)}_{F_A}$ is a primitive form for the tuple $(\mathcal{H}^{(0)}_{F_A}, \nabla, K_{F_A})$ with the minimal exponent $r=1$. Then by Theorem 7.5 of [\textit{K. Saito} and \textit{A. Takahashi}, Proc. Symp. Pure Math. 78, 31--48 (2008; Zbl 1161.32013)], we obtain a Frobenius structure on $M$ (Corollary 3.2). In \S4, this structure is shown to satisfy the conditions of Theorem 3.1 of [\textit{Y. Ishibashi} et al., J. Reine Angew. Math. 702, 143--171 (2015; Zbl 1323.53094)] (Theorem 4.1). As a consequence, the following Corollary is obtained. \textbf{Corollary 4.5}. There exists an isomorophism of Frobenius manifolds between the one obtained from the Gromov-Witten theory for $\mathbb{P}_A^1$ and the one constructed from the pair $(f_a, \zeta_A)$. The authors remark, that this result is already known for a wide class of $A$. But previous proofs did not determine primitive forms. This paper concludes to construct the primitive form $\zeta_A$ from the Frobenius structure on $M$ in \S5 (with aid of data collected in the appendix).