Let $\Delta$ be a simplicial fan in the lattice $N\cong\mathbb Z^n$ with $d$ edges and $\{u_1,\cdots,u_d\}$ are the primitive vectors along the edges. Let $A$ be the divisor class group of the corresponding toric variety $X(\Delta)$. If $\{u_1,\cdots,u_d\}$ spans $N_\mathbb R$ then we have an exact sequence $$ 0\to M\to \bigoplus_{i=1}^d\mathbb ZD_{u_i}\to A\to0$$ where $D_{u_i}$ is the prime divisor corresponding to the edge generated by $u_i$ and $M$ is the dual lattice to $N$.

Now $A$ is a finitey generated abelian group and hence we have $A\cong\mathbb Z^l\times H$ where $H$ is a finite group. Also since $X(\Delta)$ is simplicial we know that it has only finite abelian quotient singularities.

We say that $p\in X$ (an irreducible varoety over $\mathbb C$) is a finite quotient singularity if there is a finite group $G\subseteq\text{GL}(n,\mathbb C)$ and analytic open neighbourhoods $p\in U$ in $X$ and $0\in V$ in $\mathbb C^n/G$ that $U\cong V$ as analtyic varieties and $p\mapsto 0$

Now any $p\in X(\Delta)$ is in some $U_\sigma$ where $\sigma\in\Delta$. let dimension of $\sigma$ be $k$. We have $U_\sigma\cong(\mathbb C^k\times(\mathbb C^*)^{n-k})/G_\sigma$ where $G_\sigma$ is obtained as follows - Let $N'$ be the sublattice of $N$ generated by the primitive vectors along the edges in $\sigma$ and extend it to $N''$ a sublattice of finite index in $N$. Then $G_\sigma=N/N''$

Question 1 - Is there some relation between $H$ and $G_\sigma$ for all $\sigma\in\Delta$? For example is $H$ a subgroup of $G_\sigma$?

So given any point $p\in X(\Delta)$ we have a unique smallest cone $\tau$ (of dimension $r$) such that $p\in U_\tau\cong$ an open subset of $\mathbb C^r/G_\tau$. Now for every $k$ - dimensional $\sigma\ge\tau$ we know that $p\in U_\sigma\cong$ an open subset of $\mathbb C^k/G_\sigma$.

Question 2 - Is there any relation between $G_\tau$ and $G_\sigma$?

Thank you.


Question 2 is easy: Let $N_\sigma$ be the sublattice spanned by the primitive edge vectors of $\sigma$ and let $\overline N_\sigma$ be its saturation: $\overline N_\sigma=\mathbb QN_\sigma\cap N$. Then $G_\sigma=\overline N_\sigma/N_\sigma$. Now if $\sigma$ is a face of $\tau$ then $N_\sigma=\overline N_\sigma\cap N_\tau$ and therefore $$ G_\sigma\subseteq G_\tau. $$

Question 1 is a bit more subtle. I claim that if $\dim\sigma=n$ then $H$ (the torsion in the divisor class group of $X$) is isomorphic to a subgroup of $G_\sigma$ but this embedding is not canonical.

For that let $H_\sigma$ be the divisor class group of $U_\sigma$. Restriction to $U_\sigma$ yields a sequence $$ 0\to\bigoplus_{u_i\not\in\sigma}\mathbb Z D_{u_i}\overset\iota\to A\to H_\sigma\to 0 $$ Here $\iota$ is injective since all relations between the $D_{u_i}$ with $u_i\not\in\sigma$ come from invertible functions on $U_\sigma$. But these are all constant since $\sigma$ is of maximal dimension. This implies that the torsion subgroup of $A$ injects into $H_\sigma$.

The group $H_\sigma$ can also be computed as the cokernel of $$ \mathbb Z^n\cong M\to\bigoplus_{u_i\in\sigma}\mathbb ZD_{u_i}. $$ But the matrix defining this map is the transpose of the map defining the injection $N_\sigma\hookrightarrow N\cong\mathbb Z^n$. The cokernel of the latter is $G_\sigma$. Now observe that an integral $n\times n$-matrix has the same Smith normal form as its transpose. So the cokenrels are isomorphic: $H_\sigma\cong G_\sigma$, but not canonically.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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