\(F\)-singularities under generic linkage (Q1644997)

Let \(R=k[x_1,\ldots,x_n]\) be a polynomial ring over a field, and \(I \subseteq R\) be an equi-dimensional ideal of height \(c\), generated by elements \(f_1,\ldots,f_r\). Let \(S=R[u_{ij}]_{c \times r}\) be a polynomial extension of \(R\), and set \(g_i = \sum_{j=1}^r u_{ij}f_j\). The ideal \(J=(g_1,\ldots,g_c):_S I\) is called the first generic link of \(I\). This article studies the relation between the singularities of \(R/I\) and \(S/J\). In [Am. J. Math. 124, No. 6, 1103--1124 (2002; Zbl 1029.14016)], \textit{M. Chardin} and \textit{B. Ulrich} proved that if \(R/I\) has rational singularities in characteristic zero (resp. is F-rational in characteristic \(p>0\)), then \(S/J\) has rational singularities (resp. is F-rational). In characteristic zero, Niu extended this result by proving that the Grauert-Riemenschneider canonical sheaf of \(S/J\) coincides with \(\mathscr{J}(R,I^c) \cdot (S/J)\), where \(\mathscr{J}(R,I^c)\) is the multiplier ideal of the pair \((R,I^c)\) [\textit{W. Niu}, Am. J. Math. 136, No. 6, 1665--1691 (2014; Zbl 1312.14113)]. This recovers the aforementioned result of Chardin and Ulrich about rational singularities. Moreover, Niu obtains an inequality about log-canonical thresholds, that is, \(\mathrm{lct}(S,J) \geq\mathrm{lct}(R,I)\). This implies that, if \((R,I^c)\) is log-canonical, then so is \((S,J^c)\). Inspired by Niu's results, the authors obtain analogous achievements in characteristic \(p>0\). They prove that, if \(I\) is reduced and has a reduction which is generated by at most \(c+1\) elements, then \(\tau(\omega_{S/J}) \cong \tau(R,I^c) \cdot (S/J)\). Here, \(\tau(\omega_{S/J})\) is the parameter test submodule of \(S/J\), and \(\tau(R,I^c)\) is the test ideal of the pair \((R,I^c)\). This result for instance implies that \(S/J\) is F-rational if \(R/I\) is Cohen-Macaulay and \((R,I^c)\) is F-pure. The authors also prove a result for F-pure thresholds, the characteristic \(p>0\) analog of log-canonical thresholds. Under the assumption that \(I\) has a reduction generated by \(c\) elements, they prove that \(\mathrm{fpt}(S,J) \geq\mathrm{fpt}(R,I)\). In particular, if \((R,I^c)\) is F-pure, then so is \((S,J^c)\).