On algebraic dynamics of birational automorphisms of Calabi-Yau manifolds and the universal Coxeter groupsThanks: The author is supported by JSPS Grant-in-Aid (A) 25H00587, 25K21992 and NCTS scholar program.

Keiji Oguiso Address: Mathematical Sciences, the University of Tokyo, Meguro Komaba 3-8-1, Tokyo, Japan and National Center for Theoretical Sciences, Mathematics Division, National Taiwan University, Taipei, Taiwan Email address: oguiso@ms.u-tokyo.ac.jp

Abstract.

This is a continuation of our study of algebraic dymanics of birational automorphisms of complex Calabi-Yau manifolds in the view of primitivity and the existence of Zariski dense orbit. The universal Coxeter groups and their Coxeter elements also play important roles in our study.

1. Introduction

Unless stated otherwise, we work in the category of projective varieties defined over an uncountable algebraically closed field $k$ with a fixed embedding $k\subset{\mathbb{C}}$ . Our main results are Theorems 1.6, 1.9, which are generalizations of our previous works [Og18] and [Og19], with corrections of some arguments there, and Theorem 3.2 on the universal Coxeter group.

Let $V$ be a projective variety of dimension $d\geq 2$ and $f:V\dasharrow V$ be a birational self-map, i.e., $f\in{\rm Bir}\,(V)$ . Here and hereafter we denote by $f^{n}$ the $n$ th iteration of $f$ . Algebraic dynamics studies behaviours of $f^{n}$ under $n\to\infty$ . One of the most natural and basic questions is the existence of Zariski dense orbit:

Question 1.1.

When does $f$ have a Zariski dense orbit, i.e., when is there $x\in V(k)$ such that the orbit $\{f^{n}(x)\,|\,n\in{\mathbb{Z}}_{\geq 0}\}$ of $x$ under $f$ is Zariski dense in $V$ ? (See Section 2 for a more precise formulation.)

By Amerik and Campana [AC13, Théorème 4.1] (see also [BGR17, Theorem 1.2] and Theorem 2.1 in Section 2) and by the uncountability assumption of $k$ , we have:

Theorem 1.2.

Let $f\in{\rm Bir}\,(V)$ with ${\rm ord}\,f=\infty$ . Then $f$ has a Zariski dense orbit if and only if there is no commutative diagram

for any projective variety $W$ with $0<\dim W<\dim V$ (under the uncountability of $k$ ). Moreover, in this case, the set of $x\in V(k)$ with Zariski dense orbit $\{f^{n}(x)\}_{n\geq 0}$ is also Zariski dense in $V$ .

In particular, under the setting of Theorem 1.2, the existence of a Zariski dense orbit is invariant under a birational modification, that is, if $\nu:V^{\prime}\dasharrow V$ is a birational map, then $f\in{\rm Bir}\,(V)$ has a Zariski dense orbit if and only if $f^{\prime}:=\nu^{-1}\circ f\circ\nu$ has a Zariski dense orbit.

Remark 1.3.

Theorem 1.2 is widely open if $k$ is countable, e.g. if $k=\overline{{\mathbb{Q}}}$ . See Matsuzawa-Wang [MW24], Matsuzawa-Xie [MX25], especially Theorem 1.6 there, for important progress.

The notion of primitivity of $f\in{\rm Bir}\,(V)$ is introduced by De-Qi Zhang in his study of algebraic dynamics on higher dimensional projective varieties [Zh09] (See also [Og14]):

Definition 1.4.

$f\in{\rm Bir}\,(V)$ is said to be primitive if there is no commutaive diagram

for any projective variety $W$ with $0<\dim W<\dim V$ . In particular, ${\rm ord}\,f=\infty$ if $f$ is primitive.

Primitivity of $f$ is also invariant under a birational modification in the sense above. By Theorem 1.2, primitivity of $f$ implies that $f$ has a Zariski dense orbit under the assumption that $k$ is uncountable. We also remark that primitivity of is a kind of irreducibility in the category of biratioal equivalence classes of projective varieties.

Main objects of this paper is the existence of a Zariski dense orbit and primitivity of a birational automorphism of a Calabi-Yau manifold and a minimal Calabi-Yau variety. See Theorems 2.3, 2.4 for the reason why Calabi-Yau manifolds and minimal Calabi-Yau varieties are natural and important as target varieties in this study.

Definition 1.5.

Let $V$ be a projective variety of dimension $d$ .

(1)

$V$ is said to be a strict Calabi-Yau manifold, if $V$ is smooth, $H^{0}(V,\Omega_{V}^{j})=0$ for $0<j<d$ , $H^{0}(\Omega_{V}^{d})=k\omega_{V}$ with a nowhere vanishing regular $d$ -form $\omega_{V}$ and the underlying complex manifold $V_{{\mathbb{C}}}^{{\rm an}}$ is simply-connected.
(2)

$V$ is said to be a minimal Calabi-Yau variety if $V$ has at most ${\mathbb{Q}}$ -factorial terminal singularities, $H^{1}(V,{\mathcal{O}}_{V})=0$ and the numerical equivalence class of canonical divisor $K_{V}$ is $0$ , i.e., $K_{V}\equiv 0$ in $N^{1}(V)_{{\mathbb{Q}}}$ . We note that the last condition is equivalent to a stronger condition that ${\mathcal{O}}_{V}(mK_{V})\simeq{\mathcal{O}}_{V}$ for some positive integer $m$ ([Ka13, Theorem1], see also [Go13, Theorem 1.2]).

Our first main result is the following:

Theorem 1.6.

Let $V$ be a minimal Calabi-Yau variety of dimension $d\geq 3$ and $f\in{\rm Bir}\,(V)$ . Then $f$ is primitive and $f$ has a Zariski dense orbit, provided that the following two conditions are satisfied:

(1)

For any movable effective divisor $D$ , there are a minimal Calabi-Yau variety $V^{\prime}$ and a birational map $\nu:V^{\prime}\dasharrow V$ such that $D^{\prime}=\nu^{*}D$ is semi-ample on $V^{\prime}$ ; and
(2)

The characteristic polynomial $\Phi_{f}(x)$ of $f^{*}|_{N^{1}(V)_{{\mathbb{Q}}}}$ is either irreducible over ${\mathbb{Q}}$ or $\Phi_{f}(x)=\varphi(x)(x+a)$ where $\varphi(x)\in{\mathbb{Q}}[x]$ is irreducible of degree $\geq 2$ and $a\in{\mathbb{Q}}_{>0}$ .

Remark 1.7.

(1)

Theorem 1.6 is a generalization of [Og18, Theorem 1.2] with correction of arguments [Og18, Claim 2.3 and after that]. We also remark that all results of [KP24] based on the proof of [Og18, Theorem 1.2] are valid under our correction. Our new argument closely follows [LB19, Section 4], which shows that $f\in{\rm Bir}\,(V)$ is primitive if the first dynamical degree $d_{1}(f)>1$ when $V$ is a projective hyperkähler manifold. However, contrary to the case of hyperkähler manifolds, there is a strict Calabi-Yau threefold $V$ with automorphism $f\in{\rm Aut}\,(V)$ with $d_{1}(f)>1$ but $f$ is not primitive and $f$ has no Zariski dense orbit (See [Og25, Example 1, Theorem 1.2]).
(2)

The condition (1) in Theorem 1.6 is unconditional if log MMP and log abundance theorem hold in dimension $d$ . Note that, since $D$ in (1) is effective and movable, under log MMP for $(V,\epsilon D)$ ( $\epsilon$ is a small positive rational number), all extremal contractions involved are small contractions. Thus, under the assumption of log MMP, the output $(V^{\prime},\epsilon D^{\prime})$ is obtained by a finite sequence flops under which ${\mathbb{Q}}$ -factorial terminal condition and numerically triviality of the canonical class are preserved and $\epsilon D^{\prime}\equiv K_{V^{\prime}}+\epsilon D^{\prime}$ becomes nef. Then $mD^{\prime}$ is Cartier and free for some positive integer $m$ if the log abundance conjecture holds for $V^{\prime}$ . See [KM98] and [Ka24] for basic notions and facts on log MMP.
(3)

The condition (2) is also invariant under a birational map $\nu:V\dasharrow V^{\prime\prime}$ between minimal Calabi-Yau varieties. Since $\nu$ is isomorphic in codimension one as $V$ and $V^{\prime\prime}$ have only ${\mathbb{Q}}$ -factorial terminal singularities with numerically trivial canonical class (see e.g. [Ka08, Page 240, in the proof]), $\nu^{*}:N^{1}(V^{\prime\prime})_{{\mathbb{Q}}}\simeq N^{1}(V)_{{\mathbb{Q}}}$ is a well-defined isomorphism of ${\mathbb{Q}}$ -vector spaces and thus $\Phi_{f}(x)=\Phi_{f^{\prime\prime}}(x)$ where $f^{\prime\prime}:=\nu\circ f\circ\nu^{-1}$ . Here we also note that $(f^{\prime\prime})^{*}=(\nu^{*})^{-1}\circ f^{*}\circ\nu^{*}$ as again $V$ and $V^{\prime\prime}$ are ${\mathbb{Q}}$ -factorial and $f$ and $\nu$ are isomorphic in codimension one.

We shall show Theorem 1.6 in Section 4.

Definition 1.8.

Let

V:=V_{d}:=(2,2,\ldots,2)\subset({\mathbb{P}}^{1})^{d+1}

be a general hypersurface of multi-degree $(2,2,\ldots,2)$ in $({\mathbb{P}}^{1})^{d+1}$ with $d\geq 3$ . More precisely, the word general here requires the three Zariski dense open conditions under $d\geq 3$ ([CO15, Theorem 1.3, and its proof]): $V$ is smooth, ${\rm Aut}\,(V)=\{{\rm Id}_{V}\}$ and the $d+1$ projections $p_{k}:V\to P_{k}$ ( $1\leq k\leq d+1$ ) contract no divisor on $V$ . Here $P_{k}$ is the product $({\mathbb{P}}^{1})^{d}$ obtained by removing the $k$ -th factor of $({\mathbb{P}}^{1})^{d+1}$ and $p_{k}$ is the natural projection. $V$ is a $d$ -dimensional strict Calabi-Yau manifold with $\rho(V)=d+1$ . We call $V$ a Calabi-Yau manifold of Wehler type.

Our second main result is the following:

Theorem 1.9.

(1)

Every Calabi-Yau manifold of Wehler type $V_{d}$ ( $d\geq 3$ ) has a primitive birational automorphism, hence, a birational automorphism with Zariski dense orbit.
(2)

Let $X$ be a minimal Calabi-Yau variety of $\dim X=3$ and of $\rho(X)=2$ . Then the following three statements are equivalent for $f\in{\rm Bir}\,(X)$ :

(i) ${\rm ord}\,f=\infty$ (ii) $f$ is primitive (iii) $Z_{f}\not=\emptyset$ .

Moreover, (2) holds also over $k=\overline{{\mathbb{Q}}}$ .

Theorem 1.9 is an application of Theorem 1.6.

Remark 1.10.

(1)

Algebraic dynamics of $f\in{\rm Bir}\,(V)$ of a minimal Calabi-Yau variety $V$ is interesting only when $\rho(V)\geq 2$ , as ${\rm Bir}\,(V)={\rm Aut}\,(V)$ and it is a finite group if $\rho(V)=1$ (See Section 6). There are strict Calabi-Yau manifolds with $\rho(V)=2$ and with $f\in{\rm Bir}\,(V)$ of infinite order for any dimension $d\geq 3$ . (See e.g. [Og18, Theorem 1.4]).
(2)

We will present two proofs of Theorem 1.9 (1). Both proofs are based on some remarkable properties of the universal coxeter group (Theorem 3.2). The first proof is a generalization of [Og19, Theorem 8.1] with some correction of our previous argument [Og19, Page 1395, last 9 lines of the proof of Lemma 8.5]. In this proof, $f$ is implicit. Our second proof uses the Coxeter element of the universal Coxeter group (Section 3) and provides an explicit $f\in{\rm Bir}\,(V)$ . This answers a quetion posed by Professor Shigeharu Takayama in his seminar.

We will show Theorem 1.9 (1) in Section 5 and Theorem 1.9 (2) in Section 6.

2. Preliminaries on algebraic dynamics

In this section, we recall some notions and known results on algebraic dynamics relevant to our main results.

2.1. Zariski dense orbit conjecture

Let $k$ be an algebraically closed field. In this subsection, we do not assume uncountability of $k$ and also allow $k$ to be of positive characteristic. Let $V$ be a projective variety and $f:V\dasharrow V$ be a dominant rational self-map of $V$ over $k$ . Here and hereafter we denote the indeterminacy locus of $f$ by $I(f)\subset V$ and also denote

V_{f}:=V_{f}(k):=\{x\in V(k)\,|\,f^{n}(x)\not\in I(f)\,\,\forall n\in{\mathbb{Z}}_{\geq 0}\},

Z_{f}:=Z_{f}(k):=\{x\in V_{f}(k)\,|\,\overline{\{f^{n}(x)\}_{n\geq 0}}=V\},

where $\overline{A}$ is the Zariski closure of $A\subset V$ in $V$ . Note that $V_{f}$ is Zariski dense in $V$ if $k$ is uncountable. The same is true for the countable field $k=\overline{{\mathbb{Q}}}$ by Amerik [Am11, Corollary 9]. If $x\in V_{f}$ , then the points $f^{n}(x)$ are well defined points of $V$ and satisfy $f^{n+m}(x)=f^{n}(f^{m}(x))$ .

We call a closed point $x\in V(k)$ very general (resp. general) if there are countably many resp. finitely many proper closed subvarieties $V_{i}\subset V$ ( $i\in I$ ) such that $x\in V\setminus\cup_{i\in I}V_{i}$ . For instance, every point of $V_{f}$ is a very general point of $V$ . Note that the set of very general points are dense in $V$ if $k$ is uncountable, while it might be empty if $k$ is countable.

The next theorem due to [AC13, Théorème 4.1] (see also [BGR17, Theorem 1.2]) is fundamental to the existence of Zariski dense orbit:

Theorem 2.1.

Let $V$ be a projective variety and $f:V\dasharrow V$ be a dominant rational self-map. Then there exist a projective variety $W$ (possibly ${\rm Spec}\,k$ ) and a dominant rational map $\varphi:V\dasharrow W$ such that $\varphi\circ f=\varphi$ and $\varphi^{-1}(\varphi(x))=\overline{\{f^{n}(x)\}_{n\geq 0}}$ for very general points $x\in V(k)$ within $V_{f}$ .

Remark 2.2.

(1)

Theorem 1.2 in Introduction is a direct consequence of Theorem 2.1 when $k$ is uncountable. Thus primitivity of $f$ implies $Z_{f}\not=$ under the same assumption on $k$ .
(2)

Primitivity is stronger than the existence of Zariski dense orbit even over ${\mathbb{C}}$ , as the following simple example shows. Let $E$ , $F$ be two elliptic curves defined over ${\mathbb{C}}$ , which are not isogenous. Let $a\in E$ and $b\in F$ be non-torsion points. Then the translations $t_{(a,b)}$ and $t_{b}$ form a commutative diagram

and thus $t_{(a,b)}$ is not primitive. On the other hand, since $E$ and $F$ are not isogenous,

$\overline{\{t_{(a,b)}^{n}(x,y)\}_{n\geq 0}}=(x,y)+\overline{\{n(a,b)\}_{n\geq 0}}=E\times F,$

that is, the orbit of any $(x,y)\in E\times F$ is Zariski dense in $E\times F$ .

Let $V$ be a smooth projective variety. We denote by $\kappa(V):=\kappa(V,K_{V})$ the Kodaira dimension and by $q(V):=\dim H^{1}(V,{\mathcal{O}}_{V}^{1})$ the irregularity of $V$ . Set:

{\mathcal{V}}_{d}:=\{(0,0),\ldots,(0,d-1),(0,d),(-\infty,0),\ldots,(-\infty,d-1)\}.

Then by [CLO25, Theorem 1.7], we have:

Theorem 2.3.

Let $V$ be a smooth projective variety of $d:=\dim(V)\geq 2$ and $f\in{\rm Bir}\,(V)$ . Then the following assertions hold even over $\overline{{\mathbb{Q}}}$ :

(1) If $Z_{f}\not=\emptyset$ , then $(\kappa(V),q(V))\in{\mathcal{V}}_{d}\setminus\{(0,d-1)\}$ .

(2) Conversely, for each $(\kappa,q)\in{\mathcal{V}}_{d}\setminus\{(0,d-1)\}$ , there exist $V$ and $f\in{\rm Aut}\,(V)$ such that $(\kappa(V),q(V))=(\kappa,q)$ and $Z_{f}\not=\emptyset$ .

The following theorem due to D.-Q. Zhang [Zh09] (see also [Og14, Subsection 2.1]) is fundamental:

Theorem 2.4.

Assume that the base field $k$ is an algebraically closed field of characteristic zero. Then a smooth projective variety $V$ with a primitive $f\in{\rm Bir}\,(V)$ satisfies one of the following:

(1) $\kappa(V)=-\infty$ and $q(V)=0$ ;

(2) $\kappa(V)=0$ and $q(V)=0$ ; or

(3) birational to an abelian variety ( $\Leftrightarrow\kappa(V)=0$ and $q(V)=\dim\,V$ by [Ka81, Corollary 2]).

Assume further that in dimension $d$ both the minimal model conjecture and the weak abundance conjecture that claims $h^{0}(mK_{X})>0$ for some $m>0$ if $K_{X}$ is nef hold. Then $V$ of $\dim V=d$ in (1) and (2) are:

(1)’ a rationally connected manifold;

(2)’ birational to a minimal Calabi-Yau variety.

Theorems 2.3, 2.4 show that strict Calabi-Yau manifolds and minimal Calabi-Yau varieties are natural and interesting also as target varieties in algebraic dynamics.

2.2. Dynamical degrees of a dominant sel-fmap

In this subsection, we briefly recall the notion of dynamical degrees and their basic properties.

Definition 2.5.

Let $V$ be a smooth projective variety of $\dim V=d$ . over a field $k$ of characteristic $0$ and $H$ a very ample Cartier divisor on $V$ . Then, the $p$ th dynamical degree of a dominant rational sel-fmap $f:V\dasharrow V$ is defined by

d_{p}(f):=\lim_{n\to\infty}((f^{n})^{*}(H^{p}).H^{d-p})_{V}^{1/n}:=\lim_{n\to\infty}(q_{n}^{*}(H^{p}).p_{n}^{*}(H^{d-p}))_{V_{n}}^{1/n}\geq 1,

where $p_{n}:V_{n}\to V$ is a birational morphism from a smooth projective variety $V_{n}$ such that $q_{n}:V_{n}\to V$ is a morphism and $f^{n}=q_{n}\circ p_{n}^{-1}$ .

Dynamical degrees satisfy the following basic properties (See Dinh-Sibony [DS05] and references therein):

Theorem 2.6.

(1)

$d_{p}(f)$ ( $f\in{\rm Bir}\,(V)$ ) exists and it is independent of the choices involved. Moreover, $d_{p}(f)$ is a birational invariant in the sense that if $\nu:V\dasharrow V^{\prime}$ is a birational map, then $d_{p}(\nu\circ f\circ\nu^{-1})=d_{p}(f)$ .
(2)

For a surjective morphism $f:V\to V$ over ${\mathbb{C}}$ , $d_{p}(f)$ is the spectral radius $r_{p}(f)$ of $f^{*}|N^{p}(V)$ and the topological entropy $h_{{\rm top}}(f)\geq 0$ satisfies:

$(0\leq)h_{{\rm top}}(f)={\rm Max}_{p}\log d_{p}(f)={\rm Max}_{p}\log r_{p}(f).$

Salem numbers play important roles in the complex and algebraic dynamics of surface automorphisms and Coxter group theory since the breakthrough works of McMullen [Mc02a], [Mc02b]. Salem numbers are also important in our study.

Definition 2.7.

A real algebraic integer $a$ is called a Salem number if $a>1$ and its Galois conjugate are $a$ , $1/a$ and some complex numbers $e_{i}$ with $|e_{i}|=1$ (possibly empty). We call the minimal monic polynomial $\varphi_{a}(x)\in{\mathbb{Z}}[x]$ of a Salem number $a$ the Salem polynomial of $a$ . We call a monic polynomial $\varphi(x)\in{\mathbb{Z}}[x]$ a generalized Salem polynomial if it is the minimal polynomial of $a$ or $-a$ for some Salem number $a$ . We note that a generalized Salem polynomial is an irreducible reciprocal polynomial of even degree.

Proposition 2.8.

Let $L=(L,B)$ be an integral hyperbolic lattice of rank $r+1$ , that is, $L$ is a free ${\mathbb{Z}}$ -module of rank $r+1$ and $B:L\times L\to{\mathbb{Z}}$ is a symmetric bilinear form of signature $(1,r)$ . Then the characteristic polynomial of $f\in{\rm O}(L,B)$ is the product of at most one generalized Salem polynomial counted with multiplicities and cyclotomic polynomials (possibly $1$ ).

Proof.

Set $L_{{\mathbb{R}}}=L\otimes_{{\mathbb{Z}}}{\mathbb{R}}$ and $B_{{\mathbb{R}}}=B\otimes_{{\mathbb{Z}}}{\mathbb{R}}$ . Then the set $\{x\in L_{{\mathbb{R}}}|B_{{\mathbb{R}}}(x,x)>0\}\subset L_{{\mathbb{R}}}$ consists of two connected components, say $P$ and $-P$ . Then $f(P)=P$ or $f(P)=-P$ . In the first case, the characteristic polynomial of $f$ is the product of at most one Salem polynomial counted with multiplicities and cyclotomic polynomials (including $1$ ) (See eg. [Mc02b, Proof of Proposition 7.1]). In the second case, we have $(-f)(P)=P$ and the result follows from the first case applied for $-f$ . ∎

3. Universal Coxeter group and Coxeter element

The universal Coxeter group and its Coxeter element play crucial role in our proof of Theorem 1.9. Our main result of this section is Theorem 3.2. Besides its applicability for Theorem 1.9, Theorem 3.2 has its own interest and will be useful in other study.

Let $M=(m_{ij})$ be a symmetric $(r+1)\times(r+1)$ -matrix with $m_{ij}\in{\mathbb{Z}}_{\geq 2}\cup\{\infty\}$ and $m_{ii}=1$ . We define the group $G=G(M)$ by

G:=G(M):=\langle s_{1},s_{2},\cdots,s_{r+1}\,|\,(s_{i}s_{j})^{m_{ij}}=1\rangle.

The pair $(G(M),S:=\{s_{i}\}_{i=1}^{r+1})$ , denoted simply by $G$ , is called the Coxeter group with the Coxeter system $(S,M)$ . We call the element

w:=w_{G}:=s_{1}s_{2}\cdots s_{r+1}\in G

the Coxeter element of $G$ . It is known that $s_{\sigma(1)}s_{\sigma(2)}\cdots s_{\sigma(r+1)}$ ( $\sigma\in S_{r+1}$ ) is conjugate to $w$ if the Coxeter graph associated to $G$ is connected and it contains no circle (See e.g. [Hum90, 5.1, 8.4]).

We define the symmetric bilinear form $B_{G}$ on $V_{G}:={\mathbb{R}}^{r+1}=\oplus_{j=1}^{r+1}{\mathbb{R}}e_{j}$ by $B(e_{i},e_{j})=\cos(\pi/m_{ij})$ for all $1\leq i,j\leq r+1$ , so that $B(e_{i},e_{i})=-1$ in our sign convention. We denote by ${\rm O}(V_{G},B_{G})$ the real orthogonal group of $V_{G}$ with respect to the bilinear form $B_{G}$ . Then we can define the group homomorphism $\rho:G\to{\rm O}(V_{G},B_{G})$ by $s_{i}\mapsto\rho(s_{i})$ , where

\rho(s_{i})(x):=x+2B(x,e_{i})e_{i}\,\,(x\in V_{G}).

Our sign convention here fits well with our purpose. It is wellknown that $\rho$ is a well-defined group homomorphism and it is injective (See e.g. [Hum90, Page 113, Corollary]). We call the representation $\rho$ the geometric representation of the Coxeter group $G$ .

Benoist and de la Harpe [BH04, Théorème, Page 1358] proved the following:

Theorem 3.1.

$G$ be the Coxeter group. Assume that the bilinear form $B_{G}$ on $V_{G}$ is non-degenerate and indefinite. Then the image $\rho(G)$ of the geometric representation $\rho:G\to{\rm O}(V_{G},B_{G})$ is Zariski dense in the real algebraic group ${\rm O}(V_{G},B_{G})$ .

The universal Coxeter group $({\rm UC}(r+1),\{s_{i}\}_{i=1}^{r+1})$ of rank $r+1$ is the Coxter group with $m_{ij}=\infty$ for all $1\leq i\not=j\leq r+1$ , i.e.,

{\rm UC}(r+1)=\langle s_{1}\rangle*\langle s_{2}\rangle*\cdots*\langle s_{r+1}\rangle\simeq*^{r+1}{\mathbb{Z}}/2.

Let $L:=\oplus_{j=1}^{r+1}{\mathbb{Z}}e_{j}$ . Then the bilinear form $B:=B_{{\rm UC}(r+1)}$ of ${\rm UC}(r+1)$ satisfies $B(e_{i},e_{i})=-1$ and $B(e_{i},e_{j})=1$ ( $i\not=j$ ). Therefore, the geometric representation ${\rm UC}(r+1)$ is defined over ${\mathbb{Z}}$ , that is,

\rho:{\rm UC}(r+1)\to{\rm O}(L,B),

where ${\rm O}(L,B)$ is the integral linear orthogonal group of $L\simeq{\mathbb{Z}}^{r+1}$ with respect to the integral bilinear form $B$ on $L$ , and we may regard ${\rm UC}(r+1)$ as a subgroup of ${\rm O}(L,B)$ . The signature of $B$ is $(1,r)$ if $r\geq 2$ and therefore $(L,B)$ is a hyperbolic lattice of rank $r+1$ for $r\geq 2$ . (See e.g. [CO15, 2.2.2] for the proof.)

We will use the following theorem (Theorem 3.2) in our proof of Theorem 1.9. To my best knowledge, there is even no literature which states Theorem 3.2. Since the proof is fairly elementary, we shall also give a proof of Theorem 3.2.

Theorem 3.2.

Let $r\geq 2$ and regard ${\rm UC}(r+1)\subset{\rm O}(L,B)$ via the geometric representation $\rho$ . Let $\Phi_{g}(x)$ be the characteristic polynomial of $g\in{\rm UC}(r+1)$ regarded as a linear transformation of $L$ . Then:

(1)

There are infinitely many $g\in{\rm UC}(r+1)$ such that $\Phi_{g}(x)$ is a generalized Salem polynomial of degree $r+1$ when $r$ is odd and there are infinitely many $g\in{\rm UC}(r+1)$ such that $\Phi_{g}(x)$ is the product of a generalized Salem polynomial of degree $r$ and $x+1$ when $r$ is even.
(2)

The characteristic polynomial $\Phi_{w}(x)$ of the Coxeter element $w$ of ${\rm UC}(r+1)$ is a Salem polynomial of degree $r+1$ if $r$ is odd and the product of a Salem polynomial of degree $r$ and $x+1$ .

Proof.

First we show the assertion (1). Our proof of (1) is a modification of the proof of [EOY16, Theorem 4.1].

Set $L_{{\mathbb{R}}}:=L\otimes_{{\mathbb{Z}}}{\mathbb{R}}\simeq{\mathbb{R}}^{1+r}$ and $B_{{\mathbb{R}}}=B\otimes_{{\mathbb{Z}}}{\mathbb{R}}$ . Since $B_{{\mathbb{R}}}$ is of signature $(1,r)$ . we may choose a real basis

\langle v_{1},v_{2},v_{3},\cdots,v_{r+1}\rangle

of $L_{{\mathbb{R}}}$ under which $B_{{\mathbb{R}}}$ is represented by the matrix

Q={\rm diag}\,(1,-1,-1,\cdots,-1).

We identify ${\mathbb{R}}$ -linear maps of $L_{{\mathbb{R}}}$ and their matrix representations via this real basis. We set $k=(r+1)/2$ when $r$ is odd and $k=r/2$ when $r$ is even. According to $r$ is odd or even, we consider the ${\mathbb{R}}$ -linear map of $L_{{\mathbb{R}}}$ given by the matrix

M=P\oplus R_{2}\oplus\cdots\oplus R_{k}\,\,{\rm or}\,\,M=P\oplus R_{2}\oplus\cdots\oplus R_{k}\oplus(-1)

where

P=\left(\begin{array}[]{rr}\sqrt{2}&1\\ 1&{\sqrt{2}}\end{array}\right)\,\,,\,\,R_{i}=\left(\begin{array}[]{rr}\cos 2\pi\sqrt{2}&-\sin 2\pi\sqrt{2}\\ \sin 2\pi\sqrt{2}&\cos 2\pi\sqrt{2}\end{array}\right)\,\,,

for all $2\leq i\leq k$ . Then $M\in{\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})$ .

Lemma 3.3.

The characteristic polynomial of $M$ has no cyclotomic factor nor Salem factor if $r$ is odd and the characteristic polynomial of $M$ has no cyclotomic factor nor Salem factor except $x+1$ if $r$ is even.

Proof.

By the shape of $M$ , the eigenvalues of $M$ is not an algebraic integer, except $-1$ when $r$ is even. ∎

Lemma 3.4.

There are only finitely many cyclotomic polynomials of degree $\leq r+1$ .

Proof.

(cf. [EOY16, proof of Lemma 4.4]) Let $\varphi(m)=|({\mathbb{Z}}/m{\mathbb{Z}})^{\times}|=[{\mathbb{Q}}(\zeta_{m}):{\mathbb{Q}}]$ be the Euler function. Since the positive integers $m$ with $\varphi(m)\leq r+1$ is finite by the first equality, the result follows. ∎

Let

u_{1}(x)=x+1\,\,,\,\,u_{2}(x):=x-1\,\,,\cdots\,\,,\,\,u_{N}(x)

be the cyclotomic polynomials in ${\mathbb{Q}}[x]$ of degree $\leq r+1$ , where $N$ is the cardinarity of the cyclotomic polynomials of degree $\leq r+1$ .

Let $P(r+1)\subset{\mathbb{R}}[x]$ be the set of monic polynomials of degree $r+1$ . Then $P(r+1)$ is identified with ${\mathbb{A}}_{{\mathbb{R}}}^{r+1}({\mathbb{R}})$ of the real affine space ${\mathbb{A}}_{{\mathbb{R}}}^{r+1}$ . The map

{\rm char}:{\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})({\mathbb{R}})\to P(r+1)\,\,;\,\,h\mapsto\Phi_{h}(x):={\rm det}(xI_{r+1}-h)\,\,

is given by a morphism ${\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})\to{\mathbb{A}}_{{\mathbb{R}}}^{r+1}$ of affine varieties over ${\mathbb{R}}$ .

Let $P_{i}:=\{p(x)\in P(r+1)\,\,|\,\,u_{i}(x)|p(x)\}$ ( $1\leq i\leq N$ ). Then $P_{i}$ are proper closed algebraic subsets of $P(r+1)$ . According to $r$ is odd or even, we define

Q(r+1):=\cup_{i=1}^{N}P_{i}\,\,,\,\,Q(r+1):=\cup_{i=2}^{N}P_{i}.

Then $Q(r+1)$ is a closed algebraic subset of $P(r+1)$ and thus ${\rm char}^{-1}(Q(r+1))$ is a closed algebraic subset of ${\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})({\mathbb{R}})$ . Moreover ${\rm char}^{-1}(Q(r+1))\not={\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})({\mathbb{R}})$ as $M\in{\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})({\mathbb{R}})$ in Lemma 3.3 does not belongs to ${\rm char}^{-1}(Q(r+1))$ . Hence ${\rm char}^{-1}(Q(r+1))$ is a proper closed algebraic subset of ${\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})({\mathbb{R}})$ .

Let $g\in{\rm UC}(r+1)$ . Then the characteristic polynomial $\Phi_{g}(x)$ is in ${\mathbb{Z}}[x]$ , monic and of degree $r+1$ . Moreover, $g$ preserves $B$ . Thus, by Proposition 2.8, $\Phi_{g}(x)$ is the product of cyclotomic polynomials and at most one generalized Salem polynomial counted with multiplicities. Hence if the assertion (1) would not hold, then $g\in{\rm char}^{-1}(Q(r+1))$ for all $g\in{\rm UC}(r+1)$ . Thus the Zariski closure of ${\rm UC}(r+1)$ in ${\rm O}(L_{{\mathbb{R}}},B_{{\mathbb{R}}})$ is in the proper closed algebraic subset ${\rm char}^{-1}(Q(r+1))$ , a contradiction to Theorem 3.1. This completes the proof of the assertion (1) in Theorem 3.2.

Next we show the assertion (2) in Theorem 3.2. Let

w=s_{1}s_{2}\cdots s_{r+1}\in{\rm UC}(r+1)

be the Coxeter element of ${\rm UC}(r+1)$ and let $\langle e_{i}\rangle_{i=1}^{r+1}$ be the ${\mathbb{Z}}$ -basis of $L$ in the integral geometric representation of ${\rm UC}(r+1)$ .

First we observe the following:

Lemma 3.5.

With respect to the ${\mathbb{Z}}$ -basis $\langle e_{i}\rangle_{i=1}^{r+1}$ of $L$ , the matrix representation of the Coxeter element $w\in{\rm UC}(r+1)$ on $L$ is

-U\cdot U^{t}.

Here $U=(u_{kl})$ is the $(r+1)\times(r+1)$ upper triangle matrix such that $u_{kk}=1$ , $u_{kl}=-2$ for $l>k$ and $u_{kl}=0$ for $l<k$ and $U^{t}$ is the transposed matrix of $U$ .

Proof.

We can deduce this as a special case of a general result [La08, Theorem 2.9]. We may also check directly as follows.

Let $x\in L$ and write $x=\sum_{i}x_{i}e_{i}$ and $y=w(x)=\sum_{i}y_{i}e_{i}$ . Then, by

s_{j}(e_{j})=-e_{j}\,\,,\,\,s_{j}(e_{i})=e_{i}+2e_{j}\,\,(i\not=j),

Then the coefficient $z_{i}$ of $e_{i}$ in $z=\sum_{i}z_{i}e_{i}$ remains the same for $s_{j}(z)$ for any $i\not=j$ . Hence, applying $s_{r+1}$ , $s_{r}$ , $\cdots$ , $s_{j}$ inductively, we obtain that

y_{j}=-x_{j}+2\sum_{i<j}x_{i}+2\sum_{i>j}y_{i}\,\,,\,\,{\rm i.e.}\,\,,\,y_{j}-2\sum_{i>j}y_{i}=-x_{j}+2\sum_{i<j}x_{i}.

In the second formula, the left side is the $j$ th entry of $Uy$ , and the right side is the $j$ th entry of $-U^{t}x$ . Thus

Uy=-U^{t}x\quad\text{for every }x\in L

Since, $\det U=1\not=0$ , it follows that $w=-U^{-1}U^{t}$ as matrices with respect to the basis $\langle e_{i}\rangle_{i=1}^{r+1}$ . ∎

Lemma 3.6.

The characteristic polynomial $\Phi_{w}(x)$ of $w$ is

\Phi_{w}(x)=\frac{x(x+3)^{r+1}-(3x+1)^{r+1}}{x-1}.

Proof.

This is true for $r=1$ . We prove by induction of $r$ . We write the $(r+1)\times(r+1)$ matrix $U$ by $U_{r+1}$ . By the definition and by $\det U_{r+1}=1$ , we have by Lemma 3.5 that

\Phi_{w}(x)=\det(xI+U_{r+1}^{-1}U_{r+1}^{t})=\det(xU_{r+1}+U_{r+1}^{t}).

Set $D_{r+1}:=(d_{kl})=xU_{r+1}+U_{r+1}^{t}$ . Then $d_{kk}=x+1$ , $d_{kl}=-2x$ for $k<l$ and $d_{kl}=-2$ for $k>l$ . By subtracting the second line from the first line of $D_{r+1}$ and then by expanding with respect to the first line, we obtain

\det D_{r+1}=(x+3)\det D_{r}+(3x+1)\det E_{r},

where $E_{r}$ is the $r\times r$ matrix obtained by getting rid of the first line and the second colum of $D_{r+1}$ . Then, for $E_{r}$ , by subtracting the second line from the first line and then by expanding with respect to the first line, and continuing this process, we readily find that $\det E_{r}=-2(3x+1)^{r-1}$ . Substituting this and the induction hypothesis, we obtain

\Phi_{w}(x)=\det D_{r+1}=(x+3)\frac{x(x+3)^{r}-(3x+1)^{r}}{x-1}-2(3x+1)^{r}

=\frac{x(x+3)^{r+1}-(3x+1)^{r+1}}{x-1},

as claimed. ∎

Lemma 3.7.

Let $r\geq 2$ and $w$ be the Coxeter element of ${\rm UC}(r+1)$ . Then the characteristic polynomial $\Phi_{w}(x)$ of $w$ is a Salem polynomial of degree $r+1$ when $r$ is odd and $\Phi_{w}(x)$ is the product of a Salem polynomial of degree $r$ and $x+1$ when $r$ is even.

Proof.

Let $r\geq 2$ and set $f(x):=x(x+3)^{r+1}-(3x+1)^{r+1}$ . Note that $f(1)=0$ and thus $\Phi_{w}(x)=f(x)/(x-1)$ (Lemma 3.6) is certainly a monic polynomial in ${\mathbb{Z}}[x]$ . Since

(\frac{7}{5})^{r+1}\geq(\frac{7}{5})^{3}=\frac{343}{125}>2

it follows that

f(2)=2\cdot 5^{r+1}-7^{r+1}=5^{r+1}(2-(\frac{7}{5})^{r+1})<0\,\,,\,\,\lim_{x\to+\infty}f(x)=+\infty.

Thus $f(x)$ , hence $\Phi_{w}(x)$ , has a real root in $(2,\infty)$ . Hence by Proposition 2.8, $\Phi_{w}(x)$ has exactly one Salem factor. Compute that

f^{\prime}(1)=-4^{r}r\not=0

for all $r\geq 2$ ,

f(-1)=-2^{r+2}\not=0

if $r$ is odd and

f(-1)=0\,\,,\,\,f^{\prime}(-1)=-(2r+1)2^{r+1}\not=0

if $r$ is even.

By Proposition 2.8, we are now reduced to show that $f(\zeta)\not=0$ for any root of unity $\zeta\not=\pm 1$ . Assume to the contrary that $f(\zeta)=0$ for some $m$ th root of unity $\zeta$ with $m\geq 3$ . Since $f(x)\in{\mathbb{Z}}[x]$ , it follows that $f(e^{2\pi i/m})=0$ as well, as $\zeta$ and $e^{2\pi i/m}$ are Galois conjugate over ${\mathbb{Q}}$ . Hence we may and will assume that $\zeta=e^{2\pi i/m}$ and $f(\zeta)=0$ . By the shape of $f(x)$ , we have then

\zeta(\zeta+3)^{r+1}=(3\zeta+1)^{r+1}\,\,,{\rm i.e.},\,\,\zeta=(\frac{3\zeta+1}{\zeta+3})^{r+1}.

Therefore

\tau:=\frac{3\zeta+1}{\zeta+3}

is also a root of unity. Notice that the triangle with vertices $0$ , $3\zeta$ , $3\zeta+1$ and the triangle with vertices $O:=0$ , $A:=3$ , $B:=\zeta+3$ in the plane ${\mathbb{C}}$ are congruent and thus

0<\beta:={\rm arg}\,\tau=\frac{2\pi}{m}-2\alpha<\frac{2\pi}{m},

where $\alpha=\angle AOB$ . Let $C:=1$ and $D:=\zeta+1$ . Then

\overline{OD}=2\cos(\pi/m)<2=\overline{BD}

and thus

\frac{\pi}{m}-\alpha=\angle DOB>\angle DBO=\alpha\,\,,\,\,{\rm i.e.}\,\,,\,\,\frac{\pi}{m}>2\alpha.

Therefore

\frac{2\pi}{m}>{\rm arg}\,\tau=\beta=\frac{2\pi}{m}-2\alpha>\frac{2\pi}{m}-\frac{\pi}{m}=\frac{\pi}{m},

and hence

\frac{\pi}{m}<{\rm arg}\,\tau<\frac{2\pi}{m}.

However, this is impossible, as $\tau$ is a root of unity in ${\mathbb{Q}}(e^{2\pi i/m})$ so that $\tau^{2m}=1$ , hence, ${\rm arg}\,\tau=2k\pi/2m=k\pi/m$ for some integer $k$ .∎

This completes the proof of Theorem 1.9 (2). ∎

4. Proof of Theorem 1.6

In this section, we prove Theorem 1.6. Throughout this section, for a dominant rational map $\pi:X\dasharrow B$ with connected fibers between projective varieties, which we call a rational fibration, we choose a resolution $\nu:\tilde{X}\to X$ of the singularities of $X$ and $I(\pi)$ so that $\tilde{\pi}=\pi\circ\nu:\tilde{X}\to B$ is a surjective morphism with connected smooth general fibers. We denote by $\tilde{X}_{b}$ the fiber over $b\in B$ of $\tilde{\pi}$ and set $X_{b}:=\nu(\tilde{X}_{b})$ . We call a rational fibration $\pi:X\dasharrow B$ is $f$ -equivariant for $f\in{\rm Bir}\,(X)$ if there is $f_{B}\in{\rm Bir}\,(B)$ such that $\pi\circ f=f_{B}\circ\pi$ .

The following important result is due to Bianco [LB19, Proposition] (see also [Og18]):

Theorem 4.1.

Let $X$ be a projective variety and $f\in{\rm Bir}\,(X)$ . Assume that $X$ admits an $f$ -equivariant rational fibration $\pi:X\dasharrow B$ such that $\dim B<\dim X$ and $\tilde{X}_{b}$ is of general type for general $b\in B$ . Then $Z_{f}=\emptyset$ , i.e., $f$ has no Zariski dense orbit.

In what follows, $V$ is a minimal Calabi-Yau variety satisfying the conditions (1) and (2) in Theorem 1.6. Following the convention above, for a rational fibration $\pi:V\dasharrow B$ , we choose a resolution $\nu:\tilde{V}\to V$ of the singularities of $V$ and $I(\pi)$ so that $\tilde{\pi}=\pi\circ\nu:\tilde{V}\to B$ is a surjective morphism with connected smooth general fibers.

Lemma 4.2.

$V$ has no $f$ -equivariant rational fibration $\pi:V\dasharrow B$ such that $0<\dim B<\dim V$ and $\kappa(\tilde{V}_{b})=0$ for general $b\in B$ .

Proof.

Assuming to the contrary that $V$ has an $f$ -equivariant rational fibration $\pi:V\dasharrow B$ such that $0<\dim\,B<\dim\,V$ and $\kappa(\tilde{V}_{b})=0$ for general $b\in B$ , we shall derive a contradiction. Our proof of Lemma 4.2 until the end of Calim 4.3 is almost identical to the relevant part of the proof of [Og18, Theorem 1.2]. However, since we argue by contradiction, to avoid any confusion, we also repeat the identical part.

By taking a resolution of singularities of $B$ , we may and will assume that $B$ is smooth. Let $H$ be a very ample divisor on $B$ . Set $L:=\nu_{*}(\tilde{\pi}^{*}H)$ . Then $|L|$ is movable and $\pi=\Phi_{\Lambda}$ for a sublinear system $\Lambda\subset|L|$ and $\Phi_{\Lambda}$ is the rational map associated to $\Lambda$ . Then, by the assumption (1) in Theorem 1.6, there are a minimal Calabi-Yau variety $V^{\prime}$ and a birational map $g:V^{\prime}\dasharrow V$ such that $g^{*}L$ is semi-ample. So, by replacing $(V,f)$ by $(V^{\prime},f^{\prime})$ with $f^{\prime}:=g^{-1}\circ f\circ g$ , we may and will assume that $L$ is semi-ample. By Remark 1.7 (3), the characteristic polynomial of $(f^{\prime})^{*}|_{N^{1}(V^{\prime})_{{\mathbb{Q}}}}$ is the same as the characteristic polynomial of $f^{*}|_{N^{1}(V)_{{\mathbb{Q}}}}$ . Let us take a sufficiently large positive integer $m$ such that $|mL|$ is free and the morphism $\varphi_{m}:=\Phi_{|mL|}$ is the Iitaka-Kodaira fibration associated to $L$ . Set $B^{\prime}=\varphi_{m}(V)$ . Then $B^{\prime}$ is normal, projective and $\dim\,B^{\prime}=\kappa(V,L)$ , the Iitaka-Kodaira dimension of $L$ . As $\Lambda\subset|L|\subset|mL|$ , the surjective morphism $\varphi_{m}:X\to B^{\prime}$ factors through $\pi:X\dasharrow B$ , and hence, there is a dominant rational map $\rho:B^{\prime}\dasharrow B$ such that $\pi=\rho\circ\varphi_{m}$ .

Claim 4.3.

$\rho:B^{\prime}\dasharrow B$ is a birational map.

Proof.

As $\pi$ is of connected fibers and $\rho$ is dominant, it suffices to show that $\dim B=\dim B^{\prime}$ . Set $\tilde{L}:=\nu^{*}L$ . Since $\dim B^{\prime}=\kappa(V,L)=\kappa(\tilde{V},\tilde{L})$ , it suffices to show that $\dim B=\kappa(\tilde{V},\tilde{L})$ .

Let $\{E_{i}\}_{i=1}^{m}$ be the set of exceptional prime divisors of $\nu$ . Choose a positive integer $n$ such that ${\mathcal{O}}_{V}(nK_{V})\simeq{\mathcal{O}}_{V}$ . Then, in ${\rm Pic}\,(\tilde{V})$ , we have

\tilde{L}=\nu^{*}L=\tilde{\pi}^{*}H+\sum_{i=1}^{m}a_{i}E_{i}\,\,,\,\,nK_{\tilde{V}}=\sum_{i=1}^{m}b_{i}E_{i}\,\,,

where $a_{i}\geq 0$ as $\nu$ resolves $I(\pi)$ and $b_{i}>0$ as $V$ is a minimal Calabi-Yau variety. Since $\tilde{L}|_{\tilde{V}_{b}}=\sum_{i=1}^{m}a_{i}E_{i}|_{\tilde{V}_{b}}$ as $H|_{\tilde{X}_{b}}$ is trivial and $nK_{\tilde{V}_{b}}=b_{i}E_{i}|_{\tilde{V}_{b}}$ by the adjunction formula. Then there is a positive integer $M$ such that $MK_{\tilde{V}_{b}}-\tilde{L}|_{\tilde{V}_{b}}$ is linearly equivalent to an effective divisor. Hence, for general $b\in B$ , we have by $\kappa(\tilde{X}_{b})=0$ that

0\leq\kappa(\tilde{V}_{b},\tilde{L}|_{\tilde{V}_{b}})\leq\kappa(\tilde{V}_{b})=0\,\,.

Hence $\kappa(\tilde{V}_{b},\tilde{L}|_{\tilde{V}_{b}})=0$ and therefore

\dim B\leq\kappa(\tilde{V},\tilde{L})\leq\dim B+\kappa(\tilde{V}_{b},\tilde{L}|_{\tilde{V}_{b}})=\dim B\,\,.

Here the second inequality follows from [Ue75, Theorem 5.11]. Hence $\dim B=\kappa(\tilde{V},\tilde{L})$ and the result follows. ∎

By Claim 4.3, we may and will assume that $\pi:V\to B$ is an $f$ -equivariant surjective morphism given by the free complete linear system $|L|$ , by replacing $\pi:V\dasharrow B$ by $\varphi_{m}:V\to B^{\prime}$ for some $m\geq 1$ .

As in [LB19], we consider the following subspace of the ${\mathbb{Q}}$ -vector space $N^{1}(V)_{{\mathbb{Q}}}$ :

M:={\mathbb{Q}}\langle(\varphi\circ\pi)^{*}N^{1}(B)_{{\mathbb{Q}}}\,|\,\varphi\in{\rm Bir}\,(B)\rangle\subset N^{1}(V)_{{\mathbb{Q}}}.

Claim 4.4.

$0\not=M\not=N^{1}(V)_{{\mathbb{Q}}}$ . Moreover $f^{*}M=M$ .

Proof.

Since $0\not=\pi^{*}N^{1}(B)_{{\mathbb{Q}}}\subset M$ , we have $M\not=0$ . Let $\varphi\in{\rm Bir}\,(B)$ . Then there are Zariski dense subsets $U,U^{\prime}\subset B$ such that $\varphi|_{U}:U\to U^{\prime}$ is an isomorphism. Since $\pi$ is a surjective morphism, it follows that $\varphi\circ\pi$ is a surjective morphism from $\pi^{-1}(U)\subset V$ onto $U^{\prime}$ , that is, $\varphi\circ\pi$ is an almost holomorphic map from $V$ . Let $v\in N^{1}(B)_{{\mathbb{Q}}}$ and choose a positive integer $m$ and very ample general divisors $h_{1}$ and $h_{2}$ on $B$ such that $mv=h_{1}-h_{2}$ in $N^{1}(B)$ . Then

(\varphi\circ\pi)^{*}(mv)=(\varphi\circ\pi)^{*}(h_{1})-(\varphi\circ\pi)^{*}(h_{2})

in $N^{1}(V)_{{\mathbb{Q}}}$ and $((\varphi\circ\pi)^{*}(h_{i}).C)_{V}=0$ ( $i=1$ , $2$ ) for a general curve $C$ in a general fiber of almost holomorphic $\varphi\circ\pi$ . Thus $M$ contain no ample class of $V$ . Hence $M\not=N^{1}(V)_{{\mathbb{Q}}}$ . Since $\pi\circ f=f_{B}\circ\pi$ and $f$ is isomorphic in codimension one and $\pi$ is a morphism, we have

f^{*}((\varphi\circ\pi)^{*}v)=(\varphi\circ\pi\circ f)^{*}v=(\varphi\circ f_{B}\circ\pi)^{*}v.

Since $\varphi\circ f_{B}\in{\rm Bir}\,(B)$ , it follows that $f^{*}M\subset M$ . Hence $f^{*}M=M$ as $f^{*}$ is an isomorphism on $N^{1}(V)_{{\mathbb{Q}}}$ . This completes the proof. ∎

Now we are ready to derive a contradiction to complete the proof of Lemma 4.2. By the assumption (2) in Theorem 1.6, $M$ is either diminsion one or codimension one and the characteristic polynomial of $f^{*}|_{N^{1}(V)_{{\mathbb{Q}}}}$ is of the form $\varphi(x)(x+a)$ , where $\varphi(x)\in{\mathbb{Q}}[x]$ is irreducible of degree $\geq 2$ and $a\in{\mathbb{Q}}_{>0}$ .

If $\dim M=1$ , then $M$ is generated by the class of $\pi^{*}h$ where $h$ is a very ample divior on $B$ and the action of $f^{*}$ on $M$ is the multiplication by $-a<0$ . Since the divisor $\pi^{*}h$ is non-zero effective, so is $f^{*}(\pi^{*}h)$ . On the other hand, $f^{*}(\pi^{*}h)=-a\pi^{*}h$ in $N^{1}(V)_{{\mathbb{Q}}}$ . Let $H$ be a very ample divisor on $V$ . We have then

0<(f^{*}(\pi^{*}h).H^{d-1})_{V}=-a((\pi^{*}h).H^{d-1})_{V}<0,

a contradiction.

Assume that $M$ is of codimension one. By assumption (2) in Theorem 1.6, we have an $f^{*}$ -stable decomposition $N^{1}(V)_{{\mathbb{Q}}}=M\oplus{\mathbb{Q}}l$ with $f^{*}l=-al$ . Note that any element of $M$ is represented by a sum of divisors not dominating $B$ by Claim 4.4 or its proof. Thus, for any curve $C$ in a general fiber of $\pi$ , we have $(m.C)=0$ for all $m\in M$ . Thus $(l.C)\not=0$ . By replacing $l$ by $-l$ if necessary, we may and will assume that $(l.C)>0$ . Let $H$ be a very ample divisor on $V$ and write $[H]=m+cl$ with $m\in M$ and $c\in{\mathbb{Q}}$ . Then, by taking the intersection with $C$ , we have $c>0$ . On the other hand, as $C$ is general and $f^{*}H$ is movable, we have also

0\leq(f^{*}H.C)=-ac(l.C)<0,

a contradiction.

This completes the proof of Lemma 4.2. ∎

Lemma 4.5.

$Z_{f}\not=\emptyset$ for $(V,f)$ in Theorem 1.6.

Proof.

Otherwise, we have a rational dominant map $\pi:V\dasharrow{\mathbb{P}}^{1}$ such that $\pi\circ f=\pi$ by Theorem 1.2, by taking a non-constant rational function on $W$ ( $\dim W>0$ ) in Theorem 1.2. Here we used the uncountability assumption of the base field $k$ . Then $f^{*}V_{b}=V_{b}$ as divisors for general $b\in{\mathbb{P}}^{1}$ as $f$ is isomorphic in codimension one. Since $V$ is ${\mathbb{Q}}$ -factorial, it follows that $1$ is an eigenvalue of $f^{*}|_{N^{1}(V)_{{\mathbb{Q}}}}$ , a contradiction to the assumption (2) in Theorem 1.6. ∎

We are now ready to prove Theorem 1.6.

Assuming that $V$ admits an $f$ -equivariant rational dominant map $\pi:V\dasharrow B$ with $0<\dim B<\dim V$ , we shall derive a contradiction. We may and will assume that $\pi$ is a rtional fibration by taking the Stein factorization.

Choose a smooth Zariski dense open subset $B_{0}\subset B$ such that the induced morphism $\tilde{V}_{0}:=\tilde{\pi}^{-1}(B_{0})\to B_{0}$ is smooth. Consider the relative canonical model of $\tilde{V}_{0}$ over $B_{0}$ ([BCHM10]):

\varphi_{0}:\tilde{V}_{0}\dasharrow K_{0}:={\rm Proj}\oplus_{m\geq 0}\tilde{\pi}_{*}{\mathcal{O}}_{\tilde{V_{0}}}(mK_{\tilde{V_{0}}/B_{0}})\,\,.

For a projective compactification $K$ of $K_{0}$ , the rational fibration $\varphi_{0}:V_{0}\dasharrow K_{0}$ induces a rational fibration $\varphi:\tilde{V}\dasharrow K$ with a resolution $\tilde{\tilde{V}}\to\tilde{V}$ of $I(\varphi)$ . Then $\varphi$ is $f$ -equivariant and $\kappa(\tilde{\tilde{V}}_{k})=0$ for general $k\in K$ . Here $\tilde{\tilde{V}}_{k}$ is the fiber of $\tilde{\tilde{V}}\to K$ over $k\in K$ . We have $0<\dim B\leq\dim K$ by the construction. On the other hand, by Lemma 4.5 and Theorem 4.1, $\tilde{V}_{b}$ is not of general type for general $b\in B$ . Thus $\dim K_{b}<\dim V_{b}$ for general $b\in B$ . Hence $0<\dim K<\dim V$ . However, this contradicts to Lemma 4.2, as $\kappa(\tilde{\tilde{V}}_{k})=0$ for general $k\in K$ . Thus $f$ is primitive. This completes the proof of Theorem 1.6.

5. Proof of Theorem 1.9 (1)

In this section, we prove Theorem 1.9 (1).

Let $d\geq 3$ and $V$ be a $d$ -dimensional Calabi-Yau manifold of Wehler type (Definition 1.8). We will check that $V$ satisfies the condition (1) in Theorem 1.6 and $V$ has a birataional automorphism $f\in{\rm Bir}\,(V)$ satisfying the condition (2) of Theorem 1.6. We also use Theorem 5.1 below.

For the statement of Theorem 5.1, we recall:

N^{1}(V)\simeq{\rm Pic}\,V=\oplus_{k=1}^{d+1}{\mathbb{Z}}h_{k}\simeq{\mathbb{Z}}^{d+1}\,\,,\,\,\overline{{\rm Amp}}\,(V)=\oplus_{k=1}^{d+1}{\mathbb{R}}_{\geq 0}h_{k}\,\,.

Here $h_{k}$ is the pullback of the hyperplane class of ${\mathbb{P}}_{k}^{1}$ under the natural projection $V\to{\mathbb{P}}_{k}^{1}$ . Let $f\in{\rm Bir}\,(V)$ . Then, by Remark 1.7 (3), $f:V\dasharrow V$ is isomorphic in codimension one and we have a natural contravariant representation

\rho^{*}:{\rm Bir}\,(V)\to{\rm GL}\,(N^{1}(V))\simeq{\rm GL}\,({\mathbb{Z}}^{d+1})\,\,;\,\,f\mapsto f^{*}\,\,.

As $V$ is smooth, we do not need to replace $N^{1}(V)$ by $N^{1}(V)_{{\mathbb{Q}}}$ .

Let $p_{k}:V\to P_{k}$ ( $1\leq k\leq d+1$ ) be the natural projection, where $P_{k}$ is the product manifold $({\mathbb{P}}^{1})^{d}$ obtained by removing the $k$ -th factor of $({\mathbb{P}}^{1})^{d+1}$ . The morphism $p_{k}$ is of degree $2$ and the covering involution $\iota_{k}$ of $p_{k}$ is a birational automorphism of $V$ of order two. $p_{k}$ is never finite as $d\geq 3$ . Then by our definition of Wehler type (Definition 1.8), the Stein factorization $\overline{p}_{k}:V\to V_{k}$ of $p_{k}:V\to P_{k}$ is a small contraction of relative Picard number $1$ . Moreover, $\iota^{*}h_{k}+h_{k}$ is invariant under $\iota_{k}$ so that $\iota^{*}h_{k}+h_{k}=p_{k}^{*}l_{k}$ for some divisor on $P_{k}$ . In particular, $h_{k}$ is $\overline{p}_{k}$ -ample and thus $\iota_{k}^{*}h_{k}$ is $\overline{p}_{k}$ -anti-ample and therefore $\iota_{k}:V\dasharrow V$ is a flop in the sense of MMP. The following theorem is proved by [CO15, Theorems 1.3, 3.3]:

Theorem 5.1.

Let $V$ be a Calabi-Yau manifold of Wehler type of dimension $d\geq 3$ . Then:

(1)

${\rm Bir}\,(V)=\langle\iota_{1},\ldots,\iota_{d+1}\rangle$ and $({\rm Bir}\,(V),\{\iota_{k}\}_{k=1}^{d+1})$ forms a Coxeter group isomorphic to the universal Coxeter group ${\rm UC}(d+1)$ of rank $d+1$ .
(2)

The representation $\rho^{*}$ is faithful and coincides with the dual of the geometric representation of the universal Coxter group $({\rm Bir}\,(V),\{\iota_{k}\}_{k=1}^{d+1})\simeq{\rm UC}(d+1)$ : for each $1\leq j\leq d+1$ ,

$\rho^{*}(\iota_{j})(h_{j})=-h_{j}+\sum_{k\not=j}2h_{k}\,\,,\,\,\rho^{*}(\iota_{j})(h_{i})=h_{i}\,\,(i\not=j).$
(3)

For any effective movable divisor $D$ , there are a nef divisor $H$ and $g\in{\rm Bir}\,(V)$ such that $g^{*}D=H$ in $N^{1}(V)$ . In particular, $H$ is semi-ample. (See the description of the nef cone $\overline{{\rm Amp}}\,(V)$ above.)

We are ready to prove Theorem 1.9 (1). $V$ satisfies the condition (1) in Theorem 1.6 by Theorem 5.1 (3). Since $\rho^{*}$ is the dual of the geometric representation of $\rho$ by Theorem 5.1, $V$ satisfies the condition (2) in Theorem 1.6 by Theorem 3.2(1). Moreover, the image of $\iota_{d+1}\iota_{d}\cdots\iota_{1}$ under $\rho^{*}$ satisfies the condition (2) in Theorem 1.6 by Theorem 3.2(2). Note that the characteristic polynomial remains the same under taking the transposed matrix. Hence $\iota_{d+1}\iota_{d}\cdots\iota_{1}$ is a primitive birational automorphism of $V$ . This completes the proof of Theorem 1.9 (1) with an explicit primitive $f\in{\rm Bir}\,(V)$ .

6. Proof of Theorem 1.9 (2)

In this section, we prove Theorem 1.9 (2). First, we show the following general result.

Proposition 6.1.

Let $k$ be an algebraically closed field of characteristic zero. Let $V$ be a minimal Calabi-Yau variety of dimension $d\geq 3$ and $f\in{\rm Bir}\,(V)$ defined over $k$ . Set $f^{*}:=f^{*}|_{N^{1}(V)_{{\mathbb{Q}}}}$ . Then:

(1)

${\rm ord}\,f=\infty$ if and only if ${\rm ord}\,f^{*}=\infty$
(2)

Assume that $\rho(V)=1$ . Then $f\in{\rm Aut}\,(V)$ and ${\rm ord}\,f<\infty$ .
(3)

Assume that $\rho(V)=2$ . Then if ${\rm ord}\,f=\infty$ , then the characteristic polynomial $\Phi_{f}(x)\in{\mathbb{Q}}[x]$ is irreducible over ${\mathbb{Q}}$ . Moreover, $d_{1}(f)>1$ .

Proof.

Let us show (1). It suffices to show that if ${\rm ord}\,f^{*}<\infty$ , then ${\rm ord},f<\infty$ . Since $H^{0}(X,T_{X})\simeq H^{1}(X,{\mathcal{O}}_{X})^{*}=0$ for a minimal Calabi-Yau variety by [Xu20, Proposition 4.3], it follows that the identity component ${\rm Aut}\,^{0}(V)$ is trivial. Thus $f$ is also of finite order by the Fujiki-Lieberman type theorem (See e.g. [Li20, Theorem 1.4] for singular projective varieties we need).

Let us next show (2). Since $f:V\dasharrow V$ is isomorphic in codimension one and preserves the ample cone by $\rho(V)=1$ , it follows that $f\in{\rm Aut}\,(V)$ by [Ka97, Lemma 1.5]. Then $f^{*}(h)=h$ for the primitive integral ample class $h\in N^{1}(V)$ as $f\in{\rm Aut}\,(V)$ . Thus $f^{*}|_{N^{1}(V)_{{\mathbb{Q}}}}={\rm id}$ again by $\rho(V)=1$ . Hence the result follows from (1).

Let us show (3). Since $\Phi_{f}(x)\in{\mathbb{Q}}[x]$ is of degree 2, the irreducibility of $\Phi_{f}(x)$ is equivalent to the fact that $\Phi_{f}(x)$ has no rational zero. Thus, it suffices to show that $\Phi_{g}(x)$ has no rational zero for $g:=f^{2}$ . Note that ${\rm ord}\,g^{*}=\infty$ by (1).

Consider the boundary rays of $\overline{{\rm Mov}}\,(X)$ . Since $\rho(X)=2$ and $\overline{{\rm Mov}}\,(X)$ is strictly convex, there are exactly two such rays, say, ${\mathbb{R}}_{>0}v_{1}$ and ${\mathbb{R}}_{>0}v_{2}$ , and $v_{1}$ and $v_{2}$ form ${\mathbb{R}}$ -basis of $N^{1}(X)_{{\mathbb{R}}}$ . Since $g=f^{2}$ and $f$ is isomorphic in codimension one, we have $g^{*}=(f^{*})^{2}$ and. Then $g^{*}v_{1}=a_{1}v_{1}$ and $g^{*}v_{2}=a_{2}v_{2}$ for some $a_{1}>0$ and $a_{2}>0$ and $\Phi_{g}(x)=(x-a_{1})(x-a_{2})$ in ${\mathbb{R}}[x]$ .

Assume to the contray that one of $a_{i}$ is rational. Then $a_{1},a_{2}\in{\mathbb{Q}}$ .

Let $H$ be a very ample Cartier divisor on $V$ .

If $a_{1}=a_{2}$ , say, $a>0$ , then for any ample integral Weil divisor $D$ , we have $(g^{n})^{*}D=a^{n}D$ for all $n\in{\mathbb{Z}}$ , and this is again an integral Weil divisor class as $g$ and $g^{-1}$ are isomorphic in codimension one. Then

0<a^{n}(D.H^{2})_{V}=((g^{n})^{*}D.H^{2})_{V}\in{\mathbb{Z}}

for all integers $n$ . Then, by considering $n\to\pm\infty$ , we see that $a=1$ . Then $g^{*}={\rm id}$ , a contradiction.

Now we may assume that $a_{1}\not=a_{2}$ . Then we may and will assume that $a_{1}\not=1$ . Since ${\mathbb{R}}v_{1}$ is the eiganspace of $g^{*}$ with eigenvalue $a_{1}$ , which is rational and of multiplicity one, we may choose an integral Weil divisor $D$ so that ${\mathbb{R}}_{>0}v_{1}={\mathbb{R}}_{>0}[D]$ . Since $(v_{1}.H^{2})_{V}\geq 0$ and since the class $H^{2}$ is in the interior of $\overline{{\rm NE}}(X)$ , we have $(v_{1}H^{2})_{V}\not=0$ and thus $(v_{1}.H^{2})_{V}>0$ . Thefore $(D.H^{2})_{V}>0$ as well. Again as before, $(g^{n})^{*}D$ is an integral Weil divisor as a divisor. It then follows that

0<a_{1}^{n}(D.H^{2})_{V}=((g^{n})^{*}D.H^{2})_{V}\in{\mathbb{Z}}

for all integers $n$ . Then, by considering $n\to\pm\infty$ again, we see that $a_{1}=1$ , a contradiction to our choice of $a_{1}$ .

Hence both $a_{1}$ and $a_{2}$ are irrational and so are $\sqrt{a_{1}}$ and $\sqrt{a_{2}}$ . Thus $d_{1}(f)\not=1$ , as $d_{1}(f)$ is either $\sqrt{a_{1}}$ or $\sqrt{a_{2}}$ as $(f^{n})^{*}=(f^{n})^{*}$ . Since $d_{1}(f)\geq 1$ , it follows that $d_{1}(f)>1$ . This completes the proof of (3). ∎

Now we are ready to prove Theorem 1.9 (2).

Let $X$ be a $3$ -dimensional minimal Calabi-Yau variety with $\rho(X)=2$ and $f\in{\rm Bir}\,(X)$ . Since $k$ is uncountable or $k=\overline{{\mathbb{Q}}}$ , it follows that (ii) implies (iii) by Theorem 1.2 and [MX25, Theorem 1.6] when $k=\overline{{\mathbb{Q}}}$ . Note that, by Proposition 6.1 (3), we have $d_{1}(f)>1=d_{3}(f)$ if $f$ is primitive so that ${\rm ord}\,f=\infty$ . It is clear that (iii) implies (i). It remains to show that (i) implies (ii).

Assume that ${\rm ord}\,f=\infty$ . We are going to show that $f$ is primitive by checking the two conditions (1) and (2) in Theorem 1.6.

By the log abundance theorem for threefold [KMM94], $X$ satisfies the condition (1) in Theorem 1.6, and by Proposition 6.1 (3), $f$ satisfies the condition (2) in Theorem 1.6. This completes the proof of Theorem 1.9 (2).

Acknowledgements

First of all, I would like to express my thanks to Professor Shigeharu Takayama for his invitation to talk at his seminar and useful discussions there. I noticed and repaired some errors in the proof of my prvious works during preparation of my talk there. I used Chat GPT plus in our study of the universal Coxeter groups in Section 3, for which I am responsible, and the proof of Theorem 3.2 presented here is due to the author.

References

[Am11] E. Amerik,: Existence of non–periodic algebraic points for a rational self-map of infinite order, Math. Res. Lett. 18 (2011) 251–256.
[AC13] Amerik, E., Campana, F., : Fibrations Méromorphes Sur Certaines Variétés à Fibré Canonique Trivial, Pure and Applied Mathematics Quarterly, Special Issue: In honor of Fedor Bogomolov, 4 (2008) 509–545.
[BGR17] J. Bell, D. Ghioca, Z. Reichstein, : On a dynamical version of a theorem of Rosenlicht, Ann. Sc. Norm. Super. Pisa Cl. Sci. 17 (2017) 187–204.
[BH04] Y. Benoist, P. de la Harpe, : Adhérence de Zariski des groupes de Coxeter, Compositio Math. 140, (2004) 1357–1366.
[BCHM10] Birkar, C., Cascini, P., Hacon, C. D.; McKernan, J., : Existence of minimal models for varieties of log general type, J. Amer. Math. Soc. 23 (2010) 405–468.
[CO15] S. Cantat, K. Oguiso, : Birational automorphism groups and the movable cone theorem for Calabi-Yau manifolds of Wehler type via universal Coxeter groups, Amer. J. Math. 137 (2015) 1013–1044.
[CLO25] J. A. Chen, H.-Y. Lin, K. Oguiso, : On the Kawaguchi-Silverman conjecture for birational automorphisms of irregular varieties, Forum Math. 37 (2025) 1727–1741.
[DN11] T.-C. Dinh, V.-A. Nguyen, :Comparison of dynamical degrees for semi-conjugate meromorphic maps, Comment. Math. Helv. 86 (2011) 817–840.
[DS05] T.-C. Dinh, N. Sibony, : Une borne supérieure pour l’entropie topologique d’une application rationnelle, Ann. of Math. 161 (2005) 1637–1644.
[EOY16] H. Esnault, K. Oguiso, X. Yu, : Automorphisms of elliptic K3 surfaces and Salem numbers of maximal degree, Algebr. Geom. 3 (2016) 496–507.
[Go13] Y. Gongyo, : Abundance theorem for numerically trivial log canonical divisors of semi-log canonical pairs, J. Algebraic Geom. 22 (2013) 549–564.
[Hum90] J. E. Humphreys, : Reflection Groups and Coxeter Groups, Cambridge University Press, 1990.
[KP24] I. Kaur, A. Prendergast-Smith, : Remark on a theorem of Oguiso, Ann. Univ. Ferrara Sez. VII Sci. Mat. 70 (2024) 859–869.
[Ka81] Y. Kawamata, : Characterization of abelian varieties, Compositio Math. 43 (1981) 253–276.
[Ka97] Y. Kawamata, : On the cone of divisors of Calabi-Yau fiber spaces, Internat. J. Math. 8 (1997) 665–687.
[Ka08] Y. Kawamata, : Flops connect minimal models, Publ. Res. Inst. Math. Sci. 44 (2008) 419–423.
[Ka13] Y. Kawamata, : On the abundance theorem in the case of numerical Kodaira dimension zero, Amer. J. Math. 135 (2013) 115–124.
[Ka24] Y. Kawamata, : Algebraic varieties: minimal models and finite generation, Translated from the 2014 Japanese original by Chen Jiang. Cambridge Studies in Advanced Mathematics 214, Cambridge University Press, Cambridge (2024).
[KMM94] S. Keel, K. Matsuki, J. McKernan, : Log abundance theorem for threefolds, Duke Math. J. 75 (1994) 99–119.
[KM98] J. Kollár, S. Mori, : Birational geometry of algebraic varieties, volume 134 of Cambridge Tracts in Mathematics. Cambridge University Press, Cambridge, (1998).
[La08] S. Ladkani, : On the periodicity of Coxeter transformations and the non-negativity of their Euler forms, Linear Algebra Appl. 428 (2008) 742–753.
[Li20] S. Li, : Derived length of zero entropy groups acting on projective varieties in arbitrary characteristic – a remark to a paper of Dinh-Oguiso-Zhang, Internat. J. Math. 31 (2020) 2050059, 10 pp.
[LB19] F. Lo Bianco,: On the primitivity of birational transformations of irreducible holomorphic symplectic manifolds, Int. Math. Res. Not. IMRN, (1):1–32, (2019)
[MW24] Y. Matsuzawa and L. Wong, :Arithmetic degrees and Zariski dense orbits of cohomologically hyperbolic maps, Trans. Amer. Math. Soc. 377 (2024) 6311–6340.
[MX25] Y. Matsuzawa and J. Xie, :Arithmetic degree and its application to Zariski dense orbit conjecture, J. Lond. Math. Soc. 112 (2025) Paper No. e70282, 17 pp.
[Mc02a] C. T. McMullen, : Dynamics on K3 surfaces: Salem numbers and Siegel disks, J. reine angew. Math. 545 (2002) 201–233
[Mc02b] C. T. McMullen, : Coxeter groups, Salem numbers and the Hilbert metric, Publ. Math. Inst. Hautes Études Sci. 95 (2002) 151–183.
[Og14] K. Oguiso, : Some aspects of explicit birational geometry inspired by complex dynamics, Proceedings of the International Congress of Mathematicians, Seoul 2014 (Invited Lectures) Vol.II (2014) 695–721.
[Og18] K. Oguiso, : A criterion for the primitivity of a birational automorphism of a Calabi-Yau manifold and an application, Math. Res. Lett. 25 (2018) 181–198.
[Og19] K. Oguiso, : Pisot units, Salem numbers, and higher dimensional projective manifolds with primitive automorphisms of positive entropy, Int. Math. Res. Not. IMRN (2019) 1373–1400.
[Og25] K. Oguiso, : Fibered Calabi-Yau threefolds with relative automorphisms of positive entropy and c2-contractions, Rend. Circ. Mat. Palermo 74 (2025) [Paper No. 4], 16 pp.
[Tr15] Truong, T.T., : (Relative) dynamical degrees of rational maps over an algebraic closed field, arXiv:1501.01523.
[Ue75] Ueno, K., Classification theory of algebraic varieties and compact complex spaces, Lecture Notes in Mathematics, 439, Springer-Verlag, 1975.
[Xu20] J. Xu,: Homogeneous fibrations on log Calabi–Yau varieties, Manuscripta Math. 162 (2020) 389–401.
[Zh09] Zhang, D.-Q., : Dynamics of automorphisms on projective complex manifolds, J. Differential Geom. 82 (2009) 691–722.