Entropy of the composition of two spherical twists (Q6171988)

In the study of derived and, more generally, triangulated categories one often encounters examples of important endofuntors. \textit{G. Dimitrov} et al. [Contemp. Math. 621, 133--170 (2014; Zbl 1348.18017)] defined a notion of \textit{entropy} for an exact endofunctor of a triangulated category equipped with a generator. They also computed the entropy of the autoequivalence given by tensoring with a line bundle. The entropy of spherical and P-twists has been computed by \textit{G. Ouchi} [Proc. Am. Math. Soc. 148, No. 3, 1003--1014 (2020; Zbl 1441.14063)] and \textit{Y.-W. Fan} [``On entropy of P-twists'', Preprint, \url{arXiv:1801.10485}], respectively. It was conjectured by \textit{K. Kikuta} and \textit{A. Takahashi} [Int. Math. Res. Not. 2019, No. 2, 457--469 (2019; Zbl 1440.14090)] that this categorical entropy of an autoequivalence, when evaluated at 0 (categorical entropy \(h_t\) depends on an extra parameter \(t\)), equals the logarithm of the spectral radius of the induced action on the numerical Grothendieck group of the category. Several counterexamples have been given to this conjecture since then. In the present paper, the authors give some bounds on the autoequivalence given by a composition of two spherical twists with some additional assumptions. The most precise result is the following. Assume \(\mathcal{T}\) is a \(k\)-linear proper enhanced triangulated category with a split-generator and a Serre functor, and let \(E_1\) and \(E_2\) be two \(d\)-spherical objects in \(\mathcal{T}\). Denote by \(T\) the composition of spherical twists around \(E_1\) and \(E_2\). Denote by \(V^\bullet\) the space \(\mathop{\mathrm{Hom}}^\bullet_{\mathcal{T}}(E_2, E_1)\) and assume further that \(\mathop{\mathrm{Ext}}^d_{D^b(k)}(V, V)=0\). If \(\dim V \leq 2\), then \(h_0(T)=0\). If \(\dim V \geq 3\), an explicit formula for \(h_0(T)\) is given (see Theorem~1.0.2). Based on this, the authors construct yet another counterexample to the spectral radius conjecture. There are also some results on \(h_t(T)\), which an interesting reader will find in the paper (see, for instance, Theorems~4.0.1 and~4.0.4).