A \((2,3)\)-complete intersection fourfold with no decomposition of the diagonal (Q6155359)

A variety \(X\) is stably rational if \(X\times\mathbb{P}^n\) is rational for some \(n\) and \(X\) is retract rational if the identity map on \(X\) factors rationally through a projective space. Stably rational varieties are retract rational. Over algebraically closed fields, it is a major open question whether there exist stably rational varieties that are not retract rational. Let us focus on complete intersection Fano fourfolds in projective spaces, and denote a complete intersection of hypersurfaces of degree \(d_1,\dots,d_n\) in \(\mathbb{P}^{n+4}\) by \((d_1,\dots,d_n)\). Without loss of generality, we ask \(d_i\geq2\). Then all the possible \((d_1,\dots,d_n)\) in the Fano range are: \[(2),(3),(4),(5),(2,2),(2,3),(2,4),(3,3),(2,2,3),(2,2,2,2).\] The rationality of very general cubic fourfolds is famously unknown. The cases \((2), (2,2)\) are known to be rational. All the other cases are known to be stably irrational, see [\textit{J. Nicaise} and \textit{J. C. Ottem}, Duke Math. J. 171, No. 15, 3023--3075 (2022; Zbl 1509.14121)]. Furthermore, except for \((2,3)\)-complete intersection, the fourfolds are known to be non-retract-rational. This paper under review shows that a very general \((2,3)\)-complete intersection is not retract rational, resolving this last case. The author finds singular \((2,3)\)-complete intersection fourfolds which do not admit a decomposition of the diagonal. More precisely, the example is a complete intersection \(X=C\cap Q\) of a cubic hypersurface \(C\subset\mathbb{P}^6\) with \(Q=V(x_3x_6-x_4x_5)\subset\mathbb{P}^6\), a cone over \(\mathbb{P}^1\times\mathbb{P}^1\). Let us blow up \(Q\) along the vertex plane, denote the exceptional divisor by \(E\). Let us denote the strict transform of \(X\) by \(X'\) and let \(\tau\colon\widetilde{X}\to X'\) be an odd degree alteration of \(X'\). Suppose \(X\) admits a decomposition of the diagonal, let \(K=k(X)=k(\widetilde{X})\), then \(\widetilde{X}\) has a decomposition of the diagonal \[\delta_{\widetilde{X}}=[z]+[z']+[z'']\] with \(z\) supported on constant cycles on \(X'-E\), \(z'\) supported on \(E^{\mathrm{sm}}_K\) and \(z''\) supported on \((X')^{\mathrm{sing}}_K\). The variety \(X'\) is a quadric surface bundle over \(\mathbb{P}^1\times\mathbb{P}^1\) that is birational to the Hassett-Pirutka-Tschinkel example. This specific example has a explicit unramified Brauer class, that can be paired with the above equality. One calculates that the class pairs nonzero with the left hand side and zero with each term on the right hand side (especially \([z']\)), giving contradiction. In [\textit{J. Lange} and \textit{B. Skauliar}, ``The diagonal of (3,3) fivefolds'', \url{arXiv:2303.00562}], the non-retract-rationality for (3,3)-fivefolds is solved. There a new cycle theoretic criterion for non-existence of decomposition applied in [\textit{N. Pavic} and \textit{S. Schreieder}, ``The diagonal of quartic fivefolds'', \url{arXiv:2106.04539}].