A non-vanishing theorem for \(\mathbb{Q}\)-divisors on surfaces (Q2575717)

Let \(X\) be a nonsingular complex projective variety of dimension \(n\) and let \(L\) be a \(\mathbb Q\)-divisor on \(X\). Inspired by several non-vanishing results in the literature, the authors propose the following conjecture. Suppose that: 1) \(L\) is nef and big, 2) \(K_X+L\) is nef and 3) either \(L\) is a Cartier integral divisor or \(L\) is effective; then \(H^0(K_X+\lceil L \rceil)\not= 0\), where \(\lceil L \rceil\) is the round up of \(L\). When \(L\) is an integral Cartier divisor and either \(n=2\) or \(n=3\) and \(K_X\) is nef this has been proven to be true by \textit{Y. Kawamata} [Asian J. Math. 4, 173--181 (2000; Zbl 1060.14505)]. The authors prove that if \(n=2\), \(X\) is relatively minimal and 1)--3) hold, then either \(H^0(K_X+\lceil L \rceil)\not= 0\) or \(H^0(K_X+4L_{\text{red}})\not= 0\). The second conclusion may occur when \(K_X\) is nef, and the authors show that it is hard to replace the coefficient ``4'' above with ``1''. Moreover they prove that assuming conditions 1) and 2) only, then \(H^0(K_X+\lceil L \rceil)\not= 0\) if either \(X\) is a surface of positive irregularity or \(X\) is a rational surface endowed with a relatively minimal elliptic fibration such that \(K_X+L\) is nef and big. An example shows that an extra condition such as 3) is necessary in the general case.