The equations of type II\(_1\) unprojection (Q932291)

Unprojection is a method that in certain circumstances can be used to compute the inverse of a birational projection between two Gorenstein varieties. Algebraically it can be thought of as making complicated Gorenstein rings from simpler ones. A model situation is the construction of a codimension 4 Gorenstein ring by \textit{A. R. Kustin} and \textit{M. Miller} [Trans. Am. Math. Soc. 284, 501--534 (1984; Zbl 0522.13011)], later generalised and expressed geometrically as Type I unprojection by \textit{S. A. Papadakis} and \textit{M. Reid} [J. Algebr. Geom. 13, No. 3, 563--577 (2004; Zbl 1071.14047)]. That case is easy to sketch for embedded varieties, although it applies far more generally. If \(D\subset X \subset{\mathbb A}^n\) is a Gorenstein divisor on a Gorenstein variety of codimension \(c\), then there is a rational function \(s\) on \(X\) that has a simple pole everywhere along \(D\). Adjoining \(s\) to the coordinate ring of \(X\) makes a new Gorenstein ring, the coordinate ring of a variety \(Y\) of codimension \(c+1\); \(X\) is a projection of \(Y\) (simply the elimination of \(s\)) with exceptional divisor \(D\), and so \(Y\) is called the unprojection of \(X\) and \(s\) the unprojection variable. The paper under review considers a similar case, in which \(D\) is not Gorenstein but is embedded with equations the \(2\times 2\) minors of the matrix \[ \begin{pmatrix} y_1 & \dots & y_n & zx_1 & \dots zx_n \\ x_1 & \dots & x_n & y_1 & \dots y_n \end{pmatrix}, \] a model identified by Reid (in an unpublished paper) as key to the simplest generalisations in which \(D\) fails to be Gorenstein. The author makes the additional assumption that \(X\) is a complete intersection. This is a subtype of a more general Type II projection. In an earlier paper [J. Algebr. Geom. 15, No. 3, 399--414 (2006; Zbl 1117.14016)], the author shows that the equations of such a Type II unprojection \(Y\) involve two new unprojection variables \(s_0,s_1\) and are generated by the equations of \(X\) together with \(2n\) equations that are affine linear in \(s_0,s_1\) and a single quadratic equation. The paper under review does three things. It computes general expressions for the \(2n\) linear equations of the unprojection. It also computes the quadratic equation in the case \(n=3\) (the case \(n=2\) was computed by Reid in the same unpublished paper and is reproduced in the paper under review). Finally, the case \(n=3\) is applied to construct certain Fano 3-folds predicted by Altınok whose pluri-anticanonical embedding is in codimension 4. The method is simply to take appropriate slices, predicted by a Hilbert series, of the general unprojection equations. The author uses the computer algebra system \texttt{Singular} [\textit{G.-M. Greuel, G. Pfister, H. Schönemann}, Comput. Sci. J. Mold. 5, No. 1, 3--9 (1997; Zbl 0902.14040)] to check that the resulting \(Y\) is an irreducible orbifold, from which it is easy to confirm that \(Y\) is a Fano 3-fold. The methods for computing the general equations boil down to a mass of multilinear algebra working in concrete free resolutions over the integers, occasionally with 2 inverted.