Embeddings of \(\mathrm{SL}(2, \mathbb Z)\) into the Cremona group (Q430771)

Let \(\text{Bir}({\mathbb P}^2)\) be the so-called Cremona group of the dimensional \(2\) complex projective space \({\mathbb P}^2\), i.e., the group of birational maps \(\varphi:{\mathbb P}^2\dashrightarrow {\mathbb P}^2\); such a map is defined by three homogeneous polynomials without common factors and of a same degree, denoted \(\deg(\varphi)\). As it has been proven some years ago, by the second author of the paper under review, the Special Linear group \(\text{SL}(n,{\mathbb Z})\) can not be embedded into \(\text{Bir}({\mathbb P}^2)\), when \(n\geq 4\), and for \(n=3\) that may be done only via linear maps. In the present paper the authors consider embeddings of \(\text{SL}(2,{\mathbb Z})\) into the Cremona group. Following a recent work of \textit{S. Cantat} [Ann. Math. (2) 174, No. 1, 299--340 (2011; Zbl 1233.14011)] one knows that the Cremona group naturally acts on a hyperbolic space of infinite dimension which allows to classify elements in \(\text{Bir}({\mathbb P}^2)\) as being elliptic, parabolic or hyperbolic; a birational map \(\varphi\) is of one of these three classes, respectively, if and only if the sequence which to a positive integer \(n\) associates \(\deg(\varphi^n)\) is bounded, has linear or quadratic growth or grows exponentially. By thinking of \(\text{SL}(2,{\mathbb Z})\) as a subgroup of Moebius transformations acting on the hyperbolic plane, an embedding \(\theta: \text{SL}(2,{\mathbb Z})\to \text{Bir}(\mathbb P^2)\) is said to preserve the type if elliptic (resp. parabolic; resp. hyperbolic) elements are sent to elliptic (resp. parabolic; resp. hyperbolic) elements. Analogously, an embedding \(\theta\) is said to be elliptic (resp. parabolic; resp. hyperbolic) if each element (resp. each element of infinite order) in the image \(\text{im}(\theta)\) of \(\theta\) is elliptic (resp. parabolic; resp. hyperbolic). In the paper under review the authors first construct a one parameter family \(\theta_\epsilon\) of embeddings preserving the type, \(\epsilon >0\), such that \(\theta_{\epsilon}\) and \(\theta_{\epsilon'}\) are not conjugate if \(\epsilon\epsilon'\neq 1\). Moreover; here \(\theta_1\) is conjugate to the natural embedding which to a matrix with rows \([a, b]\) and \([c, d]\) associates the monomial map \((x,y)\mapsto (x^ay^b,x^cy^d)\). Next they construct explicitly an infinite number of non conjugate elliptic, parabolic and hyperbolic embeddings. Finally, the authors also analyze the situation where an embedding as above comes from an embedding into the group of automorphism of a smooth rational surface; they construct infinitely many such embeddings in the elliptic and hyperbolic cases.