Periodic subvarieties of a projective variety under the action of a maximal rank abelian group of positive entropy (Q1654920)

The authors gives a classification of positive dimensional subvarieties of a normal variety $X$ of dimension $n$ which are periodic by an abelian group $G$ of rank $n-1$ acting on $X$ with positive entropy. The latter condition means that for $g\in G\backslash \{e\}$, the entropy of $g$ is positive. \par By \textit{T.-C. Dinh} and \textit{N. Sibony}'s result [Duke Math. J. 123, No. 2, 311--328 (2004; Zbl 1065.32012)], the rank $n-1$ is maximal possible. The third author proved in a previous paper [\textit{D.-Q. Zhang}, Trans. Am. Math. Soc. 368, No. 12, 8849--8872 (2016; Zbl 1369.32012)] that if such an above $X$ is not rationally connected and it has no subvariety of positive dimensional periodic under $G$, then after replacing $G$ by a finite-index subgroup, $X$ is $G$-equivariant birational to a quotient of an abelian variety $T$ by a finite group $F$. Moreover, outside a codimension 2 subset of $T$, the group $F$ acts freely. Such a $T/F$ is called quasi-étale torus quotient by the authors. Therefore, the current paper is a step forward in the direction of finding obstructions for a normal variety $X$ with such a group action $G$ not to be a quasi-étale torus quotient. \par Note that by Hironaka's equivariant resolution of singularities, we can assume that $X$ is smooth. (Entropy is a birational invariant for regular morphisms, as being well known for smooth varieties, and as shown by the authors in the current paper for general normal varieties. Many other assumptions and conclusions in results in the paper, as well as their proofs, are also birational in nature.) \par Here is one typical result in the paper. Theorem: Under some assumptions, if $Z$ is a positive dimensional irreducible subvariety of $X$ periodic under $G$, then either $Z$ is uniruled or $Z$ is fixed point wise by a finite index subgroup of $G$. Another result concerns the case where $X$ has a non-uniruled divisor periodic under $G$. Some new consequences for the case $X$ is of dimension 2 or dimension 3 (using minimal model program) are deduced. \par The key lemma for the results of the paper is the following for group actions on singular varieties, improved upon a result in [Zhang, loc. cit.] (and obtained by Dinh-Sibony [loc. cit.] in the smooth case): There are nef $\mathbb{R}$-Cartier divisors $L_i$ ($1\leq i\leq n$) such that for any $g\in G$ there is a positive number $\lambda_i(g)$ for which $g^*(L_i)$ is numerically equivalent to $\lambda_i(g) L_i$. Moreover, $L_1.L_2\ldots L_n>0$ and $A:= L_1+\ldots +L_n$ is big and nef. \par One other helpful observation is that if a variety $Z$ is fixed by an automorphism $g$, then in cohomology (or numerically): $g_*(Z)=Z$. (Note that this fact is no longer true if $g$ is not an automorphism.) Then, a result in the cited paper of the third author characterises all subvarieties $Z$ of positive dimension periodic under $G$: those varieties are exactly those belonging to $Null(A)$, that is those $Z$ for which $A^{\dim(Z)}.{Z}=0$. Hence, it remains to determine $Null(A)$. To this end, the authors use birational geometry and minima model program extensively. For example, to deal with the case in Theorem where $Z$ is non-uniruled, the authors use Lemma 3.4, which is similar to the well known fact that the group of birational selfmaps of a projective variety of general type is finite.