Hypergeometric equations and weighted projective spaces (Q763652)

For positive integers (called \textit{weights}) \(w_0,\dots, w_n\), let \(d:=\sum_{i=0}^n w_i\), and suppose that \(\mathrm{gcd}(w_0,\dots,\widehat{w_i},\dots, w_n)=1\) for all \(i\). Such a weight is called \textit{well-formed}. Consider the hypergeometric differential operator \[ H=\prod_{i=0}^n w_i^{w_i}D(D-{{1}\over{w_i}})\dots (D-{{w_i-1}\over{w_i}i}) -td^d D(D+{{1}\over{d}})\dots (D+{{d-1}\over{d}}) \] where \(D=t{{d}\over{dt}}\). Then \(H\) is singular at \(t=\lambda=(\prod w_i^{w_i})/(d^d)\). Let \(H^{\mathrm{red}}\) denote the operator obtained from \(H\) by removing from each of the two summands on the right-hand side one copy of every common factor. Then \(H^{red}\) is irreducible. Consider the variety \(Y: (\prod_{i=0}^n y_i^{w_i}=t)\,\cap\, (\sum_{i=0}^n y_i=0)\subset {\mathbb{C}}^{\times n+1}\times {\mathbb{C}}^{\times}\) and the pencil \(f=pr_2: Y\to{\mathbb{C}}^{\times}\). Here \((y_0,\dots, y_n)\) and \(t\) are coordinates of \({\mathbb{C}}^{\times n+1}\) and \({\mathbb{C}}^{\times}\), respectively. Theorem 1. Denote by \({\mathbb{V}}_{\mathbb{R}}\) the real local system of solutions of the ordinary differential equation \(H^{red}\varphi=0\). Then \({\mathbb{V}}_{\mathbb{R}}=gr_{n-1}^W R^{n-1}f_{!}{\mathbb{R}}_Y\). In particular, this naturally gives \({\mathbb{V}}_{\mathbb{R}}\) the structure of an integer local system \({\mathbb{V}}_{\mathbb{Z}}\). By Theorem 1, \({\mathbb{V}}_{\mathbb{Z}}\) supports a polarized variation of Hodge structure (PVHS). The Hodge numbers of this PVHS are computed by making use of the geometry of the anticannonical affine cone \(A=\mathrm{Spec}\bigoplus_{m\geq 0} H^0(w{\mathbb{P}}, -mK)\) over the weighted projective space \(w{\mathbb{P}}={\mathbb{P}}(w_0,\dots, w_n)\). Set \(N={\mathbb{Z}}(w_0/d,\dots, w_n/d)+{\mathbb{Z}}^{n+1}\) and let \(M=\Hom(N,{\mathbb{Z}})\). Then \(A={\mathbb{C}}^{n+1}/\mu_d\) is the quotient of \({\mathbb{C}}^{n+1}\) by \(\mu_d\) acting diagonally with weights \(w_0,\dots, w_n\). Now fix the linear form \(\ell : N\to{\mathbb{Z}}\) defined by \(\ell(\alpha_0,\dots, \alpha_n)=\sum_{i=0}^n \alpha_i\). For any integer \(a\) define \(\overline{a}\) to be the smallest representative of its class modulo \(d\). Every element \(({{a_0}\over{d}}, \dots, {{a_n}\over{d}})\in N\) has a unique representative \(({{\overline{a_0}}\over{d}},\dots, {{\overline{a_n}}\over{d}})\) mod \({\mathbb{Z}}^{n+1}\) in the unit box \([0,1)^{n+1}\subset{\mathbb{R}}^{n+1}\). This will identify the cyclic group \({\mathbb{Z}}/d=N/{\mathbb{Z}}^{n+1}\) generated by \(({{w_0}\over{d}},\dots, {{w_n}\over{d}})\) with \(N\cap[0,1)^{n+1}\). Define the \textit{age function} \(a: N/{\mathbb{Z}}^{n+1}\to [0,n]\cap {\mathbb{Z}}\) by \[ a({{a_0}\over{d}},\dots,{{a_n}\over{d}}) =\ell({{\overline{a_0}}\over{d}},\dots, {{\overline{a_n}}\over{d}}) ={{1}\over{d}}\sum_{i=0}^n \overline{a_i}. \] Denote by \((N/{\mathbb{Z}}^{n+1})^0=({\mathbb{Z}}/d)^0\) the set of classes not contained in a coordinate hyperplane: \[ ({\mathbb{Z}}/d)^0=\{0 \leq k< d\mid d\not|\,kw_i \}= \{({{\overline{kw_0}}\over{d}},\dots, {{\overline{kw_n}}\over{d}})\mid \overline{kw_i}\neq 0\}. \] For \(j=0,1,\dots, n\), define \(a_j^*=\# ({\mathbb{Z}}/d)^0\cap a^{-1}(j)\) to be the number of elements of strict age \(j\). The next main result is the computation of the Hodge numbers. Theorem 2. For \(j=1,\dots, n\), \[ rk\,{\mathcal{H}}^{n-j, j-1}(V_{\mathbb{Z}} \otimes{\mathcal{O}}_{{\mathbb{C}}^{\times}\setminus\lambda}) =h^{n-j,j-1}(H_c^{n-1} Y_t)=a_j^*. \] Brief discussion is given for Calabi-Yau hypersurfaces in weighted projective spaces \(w{\mathbb{P}}\). Theorem 3. A general hypersurface \(X_d\subset {\mathbb{P}}(w_0,\dots,w_3)\) is a well-formed \(K3\) surface if and only if it is quasismooth. Theorem 4. Fix a well-formed weighted projective space \(w{\mathbb{P}}={\mathbb{P}}(w_0,\dots, w_n)\), and let \(f: Y\to{\mathbb{C}}^{\times}\) be the pencil described by the above equation. A well-formed general hypersurface \(X_d\subset w{\mathbb{P}}\) of degree \(d=\sum w_i\) is a Calabi-Yau if and only if a general fiber \(Y_t:=f^{-1}(t)\) is birational to a Calabi-Yau variety. The proof is done on the Landau-Ginzburg model of \(Y\). Finally, the \(103\) weighted \(3\)-spaces with canonical singularities are analyzed, and shown that a general anticanonical section is not a \(K3\) surface exactly in those \(9\) cases where a generic fiber of the Landau-Ginzburg model is an elliptic surface of Kodaira dimension \(1\).