Product theorem for K-stability (Q776818)

It has been expected from differential geometric point that the product of two Fano varieties with K-stability still admits K-stability. In this article, the author gives a purely algebraic proof of this expectation. The main result is the following product result on delta invariant \[\delta(X_1\times X_2, p_1^*L_1+p_2^*L_2)=\min\{\delta(X_1, L_1), \delta(X_2, L_2)\},\] where \(L_1\) (resp. \(L_2\)) is a big line bundle on \(X_1\) (resp. \(X_2\)). The main obstruction is that, a basis type divisor of \(p_1^*L_1+p_2^*L_2\) is not necessarily of the form of \(p_1^*D_1+p_2^*D_2\), where \(D_i\) is the basis type divisor of \(L_i\). So it is not easy to directly see that the left is not smaller than the right of the above formula. For any \(c< \min\{\delta(X_1, L_1), \delta(X_2, L_2)\}\) and prime divisor \(E\) over \(X_1\times X_2\), the author creates two filtrations and cook up a basis type divisor \(\Gamma\) of \(p_1^*L_1+p_2^*L_2\) which is compatible to \(E\) and `close' to the form of \(p_1^*D_1+p_2^*D_2\), that is, one gets a basis type divisor of \(L_1\) via restricting \(\Gamma\) to the general fiber of \(p_2\). By inversion of adjunction, the author can show that \((X_1\times X_2, c\Gamma)\) is klt alone \(E\), thus proving the left is not smaller than the right of the product formula.