Dressians, tropical Grassmannians, and their rays (Q470239)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Dressians, tropical Grassmannians, and their rays
scientific article

    Statements

    Dressians, tropical Grassmannians, and their rays (English)
    0 references
    0 references
    0 references
    0 references
    12 November 2014
    0 references
    The nature of Tropical Mathematics makes linear spaces very rich and interesting. Several authors have studied tropical linear spaces in the last ten years. Among others, we have \textit{F. Ardila} and \textit{C. J. Klivans} [J. Comb. Theory, Ser. B 96, No. 1, 38--49 (2006; Zbl 1082.05021)], \textit{J. Yu} and \textit{D. S. Yuster} [``Representing tropical linear spaces by circuits'', \url{arXiv:math/0611579}], \textit{D. E. Speyer} [SIAM J. Discrete Math. 22, No. 4, 1527--1558 (2008; Zbl 1191.14076)], Herrmann, Jensen, Joswig and Sturmfels [\textit{S. Herrmann} et al., Electron. J. Comb. 16, No. 2, Research Paper R6, 26 p. (2009; Zbl 1195.14080)], \textit{F. Rincón} [Discrete Comput. Geom. 50, No. 3, 700--713 (2013; Zbl 1281.14049)] and de la Puente (2014), among others. This interesting paper continues with the task. All \((k-1)\)-dimensional tropical linear spaces in \((n-1)\)-dimensional space are contractible polyhedral complexes. Set \(k=2\). Tropical lines are well understood in any dimension. Now set \(k=3\). Prior to this paper, tropical planes were understood for \(n\leq 7\): there are one combinatorial type for \(n=4\) or \(5\), seven types for \(n=6\), and 94 types for \(n=7\). One of the byproducts of this paper is a better understanding of tropical planes for \(n=8\). The \((k-1)\)-dimensional tropical linear spaces are parameterized by the Dressian \(\mathrm{Dr}(k,n)\). The Dressian is contained in the secondary fan of the hypersimplex \(\Delta(k,n)\). It acquires a fan structure this way. Realizable \((k-1)\)-dimensional tropical linear spaces in \((n-1)\)-dimensional space are parameterized by the Tropical Grassmannian \(\mathrm{Gr}(k,n)\). It has a fan structure too. In this paper, Dressians and their close relatives Tropical Grassmannians are studied, focussing on their rays. In particular, \(\mathrm{Dr}(3,8)\) is computed. Its is a non-pure, non-simplicial 9-dimensional polyhedral fan with f-vector \((1; 15 470; 642 677; 8 892 898; 57 394 505; 194 258 750; 353 149 650; 324 404 880; 117 594 645; 113 400)\). Modulo the natural \(S_8\) symmetry, the f-vector becomes \((1; 12; 155; 1 149; 5 013; 12 737; 18 802; 14 727; 4 788; 14)\). There are \(116 962 2654\) maximal cones, among which \(113 400\) are 9-dimensional and \(116 848 865\) are 8-dimensional. Modulo the natural \(S_8\) symmetry, there are \(4 748\) maximal cones, among which \(14\) are 9-dimensional and \(4 734\) are 8-dimensional. These computations have been done using software polymake. It is difficult to double-check the result above. We have \(\mathrm{Gr}(k,n) \subseteq \mathrm{Dr}(k,n)\) but the fan structures are different. Moreover, \(\mathrm{Dr}(k,n)\) gets larger than \(\mathrm{Gr}(k,n)\), as \(n\) tends to infinity, with \(k\) fixed. Set \(1<k<n\). Let \(\mathcal F\) denote the secondary fan of the product of simplices \(\Delta_{k-1}\times \Delta_{n-k-1}\). It is proved in this paper the existence of a piecewise linear embedding \(\tau:\mathcal F\to \mathrm{Dr}(k,n)\), whose image is contained in \(\mathrm{Gr}(k,n)\). Moreover, \(\tau\) is a homeomorphism onto the image. A concept introduced in this paper is that of \textit{rigid} configuration of \(n\) points in \((k-1)\)-dimensional tropical space. These points are labeled and need not be pairwise different. Loosely speaking, rigid is opposite to generic. It is proved that any rigid configuration of \(n\) points provide a ray both in \(\mathrm{Gr}(k,n)\) and \(\mathrm{Dr}(k,n)\). Rays in \(\mathrm{Dr}(k,n)\) correspond to matroid subdivisions of the hypersimplex \(\Delta(k,n)\) that admit only the trivial coarsening. They yield the most degenerate tropical linear spaces. At the end of the paper the authors wonder whether it is feasible to compute \(\mathrm{Dr}(4,8)\). My feeling is that it is advisable to better understand the numbers above. Reviewer's remark: I think there is a mistake in the list of matrices in page 1863. The list corresponds to Figure 2. In order to get this correspondence, the last matrix must give the top left polytope in Figure 2. Thus, the last matrix must be changed to \([2,0,3; 3,2,0; 0,3,2]\), for instance.
    0 references
    0 references
    0 references
    0 references
    0 references
    Dressian
    0 references
    tropical Grassmannian
    0 references
    tropical plane
    0 references
    fan
    0 references
    ray
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references