Tutorial on Tom and Jerry: the two smoothings of the anticanonical cone over \(\mathbb{P} (1, 2, 3)\) (Q1983975)

The paper gives various examples of codimension four Gorenstein rings built using the ``Tom'' and ``Jerry'' constructions. The paper ends with two questions. In answer to the question, ``What's it all about?'', the paper says: ``A hypersurface or complete intersection is determined by the coefficients of its defining equations, so its deformations are unobstructed. The subtlety of the deformation theory in these cases is nothing to do with obstructions, but how to pass to the quotient by the appropriate equivalence relation, which involves dividing by the groupoid of local diffeomorphisms. ``The Buchsbaum-Eisenbud theorem puts codimension 3 Gorenstein ideals in the same framework: the variety is given by a skew \((2k+1)\times (2k+1)\) matrix that encodes both the defining equations and the syzygies, so that the entries of the matrix can be freely deformed. In other words, the skew matrix is a given mould, into which one can simply pour functions on the ambient space in a liquid manner. ``In contrast, one usually expects codimension \(4\) constructions to be obstructed. A typical case is the cone over a del Pezzo surface of degree \(6\), whose deformation theory has 2 components. The point of Tom and Jerry is that, in most commonly occurring cases, our variety admits a Gorenstein projection to codimension 3, with the projected variety given by the Pfaffians of a \(5 \times 5\) skew matrix; that is, the projected variety is a regular pullback from \(\operatorname{Grass}(2, 5)\) in its Plücker embedding, marked with an unprojection divisor that corresponds to a linear subspace of \(\operatorname{Grass}(2, 5)\). Every geometer must have done the easy exercise of seeing that any linear subspace of \(\operatorname{Grass}(2,4)\) (the Klein quadric) either consists of lines of \(\mathbb P^3\) passing through a point \(P\), or dually, of lines contained in a plane \(\mathbb P^2 \subseteq \mathbb P^3\). The Tom and Jerry formats discussed here answer the same question for \(\operatorname{Grass}(2, 5)\).'' In answer to the question, ``Do they do everything?'', the paper says: ``Unfortunately, no. Tom and Jerry provide two smooth components of the deformation theory, and for deformation problems entirely contained within one component or the other, they can be relied on to do everything. However, we know other cases in codimension 4 that appear not to have any useable structure of unprojection.''