Accumulation points on 3-fold canonical thresholds (Q6979213)

Canonical threshold is an invariant of a pair (variety \(X\), effective divisor on \(X\)). It is used in the works related to the Minimal Model Program. More precisely, let \(X\) be a projective variety defined over the field of complex numbers and \(B\) a \(\mathbb{Q}\)-divisor on \(X\) with coefficients in the closed interval \([0,1]\) and such that the pair \((X,B)\) is canonical. Let \(S\) be an effective \(\mathbb{Q}\)-Cartier divisor on \(X\). Then the canonical threshold of \(S\) with respect to the canonical pair \((X,B)\) is \N\[\N\mathrm{ct}(X,B;S)=\sup \{t\in \mathbb{R} \,|\, (X,B+tS) \text{ is canonical}\}.\N\]\NIn the case where \(B=0\) and \(S\) a prime effective divisor on \(X\) one writes simply \(\mathrm{ct}(X;S)\) and calls it the canonical threshold of the pair \((X,S)\). For log canonical pairs \((X,B)\) there is an analogous notion of the log canonical threshold. There were famous conjectures (now theorems) stating that, for fixed \(n=\dim X\), the set of all possible log canonical thresholds satisfies the ascending chain condition (ACC) (as a subset of \(\mathbb{R}\)) and that the set of accumulation points of log canonical thresholds in dimension \(n\) is the set of log canonical thresholds in dimension \(n-1\). The same questions can be asked about the set of canonical thresholds and, for general \(n\), remain open. Canonical thresholds, despite less known than their log canonical counterparts, are useful in Sarkisov program and in proving birational rigidity of Mori fiber spaces.\N\NDenote by \(\mathcal{T}^{can}_{n}\) the set \(\{\mathrm{ct}(X;S)\,|\, \dim X=n\}\). In the preceding work [Adv. Math. 404, Part B, Article ID 108447, 36 p. (2022; Zbl 1491.14027)], the author showed that \(\mathcal{T}^{can}_{3}\) satisfies ACC. The main result of the present paper is the theorem stating that the set of accumulation points of the set \(\mathcal{T}^{can}_{3}\) is the set \(\mathcal{T}^{can}_{2}=\{0\}\cup\{\frac{1}{k}\}_{k\in \mathbb{N}}\) of canonical thresholds in dimension \(2\). The proof uses the fact that the canonical threshold can be realized as the discrepancy of a certain divisorial contraction \(\sigma\colon Y\to X\). Such contractions were classified by Hayakawa, Kawakita and many others and turn out to be weighted blowing ups, making the problem essentially arithmetical. The present paper heavily relies on the notation and results from the above mentioned paper of \textit{J.-J. Chen}, so the interested reader can be advised to read that paper first. It should also be noted that when the results of the present paper were announced in [J. Math. Soc. Japan 77, No. 2, 325--343 (2025; Zbl 08045481)], almost simultaneously the same theorem on threefold canonical thresholds was announced by \textit{J. Han} et al. [Adv. Math. 480, Part A, Article ID 110457, 66 p. (2025; Zbl 08122408)] in a broader context. The present paper contains also a result on generalized canonical thresholds with respect to pairs \((X,B)\). Namely, let \(I\subset [0,1]\) and \(J\subset \mathbb{R}_{>0}\) be subsets satisfying ACC. Then the set of threefold canonical thresholds \(\{\mathrm{ct}(X,B;S)\}\), where the coefficients of \(B\) are from \(I\) and coefficients of \(S\) from \(J\), satisfies ACC.