In their paper "Calabi-Yau 3-folds and Moduli of Abelian Surfaces I", Gross and Popescu calculate the hodge numbers of smooth CY 3-folds obtained in the following way (see Remark 4.11 of their paper): suppose $X$ is a complete intersection CY 3-fold (i.e. dualizing sheaf is trivial) in $\mathbb P^{n+3}$ with only ODP's (ordinary double points) as singularities such that the $n$ equations $f_1,...,f_n$ which define $X$ all have the same degree $d$. Assume further that there exists a projective small resolution $\tilde{X}\rightarrow X$ of the singularities of $X$. Notice that $\tilde{X}$ is then a smooth CY 3-fold. Let $S=\mathbb C[x_0,...,x_{n+3}]/(f_1,...,f_n)$ be the homogeneous coordinate ring of $X$.

Now they claim that there is a natural isomorphism $T^1\cong \left(\frac{S^n}{\{(\partial f_1/\partial x_i,...,\partial f_n/\partial x_i)\}}\right)_d$, where the subscript $d$ refers to the degree $d$ homogeneous part of this module and $T^1\cong \operatorname{Ext}^1(\Omega_X^1,\mathcal O_X)$ is the space of infinitesimal deformations of $X$, or equivalent the tangent space to the deformation space of $X$. Why is there such a natural isomorphism?

They further state that the space of infinitesimal deformations of the small resolution $\tilde{X}$ is obtained as the kernel of the natural map $T^1 \rightarrow T^1_{\text{loc}}$, where $T^1_{\text{loc}}$ is the tangent space of the deformation space of the germ of the singular locus of $X$. Why is this the case?

Finally, they remark that this kernel is given by elements $(g_1,...,g_n)\in S^n$ representing an element of $T^1$ such that adding this as a new column to the Jacobian matrix gives a matrix which has rank $<n$ at the singular points of $X$. Why does the kernel have this description?



1 Answer 1


Since nobody has answered you, let me give you some sort of answer. It will be somewhat lacking in rigour...

First, let me frame the discussion in terms of $H^1(X, TX) \cong \mathrm{Ext}^1(\Omega^1_X, \mathcal{O}_X)$, and also write $P \equiv \mathbb{P}^{n+3}$. Start with the short exact sequence \begin{equation*} 0 \longrightarrow TX \longrightarrow TP\vert_X \stackrel{df}{\longrightarrow} \mathcal{N}_{X\vert P} \longrightarrow 0 ~. \end{equation*} Taking cohomology, this gives in part \begin{equation*} 0 \longrightarrow H^0(X, TP\vert_X) \stackrel{df}{\longrightarrow} H^0(X, \mathcal{N}_{X\vert P}) \longrightarrow H^1(X, TX) \longrightarrow 0 ~. \end{equation*} The global sections of $\mathcal{N}_{X\vert P}$ are just $n$-tuples of degree $d$ polynomials in the homogeneous coordinate ring $S$, and you can check that $H^0(X, TP\vert_X)$ is just the space of homogeneous linear polynomials. Since $df$ is represented by the matrix of first derivatives of $(f_1,\ldots, f_n)$, the specified isomorphism follows. The intuition is that polynomial deformations which are proportional to the first derivatives are realised by coordinate changes, and are therefore trivial.

This is actually a special case of an old theorem of Green and Hübsch (Commun.Math.Phys. 113 (1987) 505), which gives conditions under which counting such 'polynomial deformations' gives the value of $h^{1,2}(X)$, for $X$ a complete intersection Calabi-Yau in a product of projective spaces.

For the other parts, I don't know the details (e.g. the definition of $T^1_\mathrm{loc}$), but note that the moduli space of $\tilde X$ (the small resolution) coincides locally with the subspace of the moduli space of $X$ over which $X$ has the specified configuration of nodes. The kernel of the map $T^1 \to T^1_\mathrm{loc}$ must consist of those deformations which preserve the nodes, which can therefore be interpreted as deformations of $\tilde X$.

  • $\begingroup$ Actually we don't have $H^1(X,TX)\cong Ext^1(\Omega_X^1,\mathcal O_X)$. The deformation theory exact sequence (aka local-to-global spectral sequence) gives $0\rightarrow H^1(X,TX)\rightarrow Ext^1(\Omega_X^1,\mathcal O_X)\rightarrow H^0(X,\mathcal T^1)$. Since $\mathcal T^1$ is a skyscraper sheaf support at the ODPs and of dimension 1 at each of them, we find that $H^0(X,\mathcal T^1)=\mathbb C^k$, where $k$ is the number of ODPs. Moreover, this last map is nonzero, so those two v. spaces are not isomorphic. Moreover this last space is $T^1_loc$ as it is the deformation space of the ODPs. $\endgroup$
    – HNuer
    Jul 26, 2012 at 2:43
  • 2
    $\begingroup$ I have given a full proof of this with all of the deformation theoretic details in Section 7 of my paper "On (2,4) Complete Intersection threefolds containing an Enriques Surface" arxiv.org/abs/1210.1903. $\endgroup$
    – HNuer
    Oct 11, 2012 at 13:26

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.