Classifying smooth lattice polytopes via toric fibrations (Q2389243)

Let \(P\subset M_\mathbb{R}\cong\mathbb{R}^n\) be an \(n\)-dimensional lattice polytope represented by a finite intersection of half-spaces \[ P=\bigcap_{i=1}^rH_{\rho_i,-a_i}^+, \] where \(H_{\rho_i,-a_i}^+:=\{u\in M_\mathbb{R}\mid\left<\rho_i,u\right>\geq-a_i\}\), \(\rho_i\in N:=\mathrm{Hom}(M,\mathbb{Z})\) is a primitive inner normal, and \(a_i\in\mathbb{Z}\). We say that \(P\) is \textit{smooth} if the normal vectors generate a regular fan in \(N_\mathbb{R}\) (that is, if the corresponding toric variety is nonsingular). For each non-negative integer \(s\) we define \(P^{(s)}\) to be the polytope obtained by shifting each supporting hyperplane by \(s\), i.e. \[ P^{(s)}:=\bigcap_{i=1}^rH_{\rho_i,-a_i+s}^+. \] In particular, \(P^\circ\cap M=P^{(1)}\cap M\). Using this concept, one can define the {\(\mathbb{Q}\)-codegree} of \(P\): \[ \text{codeg}_\mathbb{Q}(P):=\inf_\mathbb{Q}\left\{\frac{a}{b}\mid (aP)^{(b)}\neq\emptyset\right\}. \] This generalises the definition of codegree introduced in [\textit{V. Batyrev} and \textit{B. Nill}, Mosc. Math. J. 7, No. 2, 195--207 (2007; Zbl 1134.52020)]. If \(u\) is a vertex of \(P\) then, after possible relabelling of the indices, we can express \(u\) as the intersection of \(n\) hyperplanes \(H_{\rho_1,-a_1},\dots,H_{\rho_n,-a_n}\). \(P\) is said to be \textit{\(s\)-spanned at \(u\)} if \(H_{\rho_1,-a_1+s}\cap\dots\cap H_{\rho_n,-a_n+s}\) defines a lattice point in \(P^{(s)}\). We call \(P\) \textit{\(s\)-spanned} if it is \(s\)-spanned at each vertex. The \textit{nef value} of \(P\) is defined to be the infimum \[ \tau(P):=\inf_\mathbb{Q}\left\{\frac{a}{b}\mid aP\text{ is \(b\)-spanned}\right\}. \] Proposition~2.2. Let \(X\) be a nonsingular projective toric variety of dimension \(n\), and let \(L\) be an ample line bundle on \(X\). Let \(P\subset M_\mathbb{R}\) be the associated smooth lattice polytope. Then \[ \tau_L=\tau(P), \] where \(\tau_L:=\min_\mathbb{R}\left\{t\mid K_X+tL\text{ is nef}\right\}\). Corollary~2.3. Let \(P\subset M_\mathbb{R}\) be an \(n\)-dimensional smooth lattice polytope. Let \(\Delta_n\) denote the standard \(n\)-simplex \(\mathrm{conv}\left\{0,e_1,\ldots,e_n\right\}\). Then: \parindent=6mm \begin{itemize}\item[(i)] \(\mathrm{codeg}(P)=\mathrm{codeg}_\mathbb{Q}(P)=\tau(P)=n+1\) if and only if \(P=\Delta_n\); \item[(ii)] if \(P\neq\Delta_n\) then \(\mathrm{codeg}_\mathbb{Q}(P)\leq\tau(P)\leq n\). With the above result in mind, a lattice polytope \(P\) is called \textit{\(\mathbb{Q}\)-normal} if \(\mathrm{codeg}_\mathbb{Q}(P)=\tau(P)\). The main result of the paper is a partial answer to a question raised in [Zbl 1134.52020]: ``Given an integer \(d\), does there exist an integer \(N(d)\) such that every lattice polytope of degree \(d\) and dimension \(\geq N(d)\) is a Cayley polytope?'' A sharp linear bound in the case of smooth \(\mathbb{Q}\)-normal lattice polytopes is established: Theorem~1.12. Let \(P\subset M_\mathbb{R}\cong\mathbb{R}^n\) be an \(n\)-dimensional smooth lattice polytope. The following are equivalent: \parindent=6mm\item[(i)] \(P\) is \(\mathbb{Q}\)-normal and \(\mathrm{codeg}(P)\geq (n+3)/2\); \item[(ii)] \(P=\mathrm{Cayley}_\Sigma(P_0,\ldots,P_k)\) is a smooth strict Cayley polytope, where \(k+1=\mathrm{codeg}(P)\) and \(k>n/2\); \item[(iii)] The toric polarised variety \((X,L)\) corresponding to \(P\) is defective, with defect \(\delta=2\mathrm{codeg}(P)-2-n\). The authors conjecture that \(\tau(P)=\mathrm{codeg}_\mathbb{Q}(P)\) always holds for smooth lattice polytopes satisfying \(\mathrm{codeg}(P)\geq (n+3)/2\), and therefore that the above result holds for all smooth polytopes.\end{itemize}