Moduli spaces of codimension-one subspaces in a linear variety and their tropicalization (Q2144314)

This paper is about the tropicalization of linear spaces defined over a field endowed with a non-Archimedean valuation, possibly trivial. Concretely, it studies the tropicalization of dual linear spaces induced by closed-embeddings into a bigger affine space for which the image is a linear space. Such closed embedding defines a \textit{model} for the tropicalization of this space, and the study of these models is relevant in tropical algebraic geometry over a valuated field, specifically for the process of embedded tropicalization of closed subsets of toric varieties. Let \(K\) be a field endowed with a non-Archimedean valuation \(\text{val}:K\xrightarrow{}\mathbb{R}\cup\{-\infty\}\) of rank at most one, and fix a pair of integers \(0\leq d<n\). Given a \((d+1)\)-dimensional subspace \(X\subset K^n\), one can assign to it a valuated matroid \(M(X)\) -- for instance, by taking the image under the map \(\text{val}\) of the Plücker vector of \(X\) --. The tropicalization of \(X\) is a subset of \(\mathbb{R}^{n-1}\) whose construction depends only on the valuated matroid \(M(X)\). Consider now a \((d+1)\)-dimensional \(K\)-linear space \(V\). Then the set \(G(d,V)\) of \(d\)-dimensional vector subspaces of \(V\) can be identified (in a non-canonical way) with its dual linear space \(V^*\). A closed embedding \(i:V\hookrightarrow{}K^n\) having image a linear subspace \(X\subset K^n\) induces a closed embedding \(j(i):V^*\hookrightarrow{}K^{\binom{n}{d}}\) whose image is also a linear subspace \(L(X)\subset K^{\binom{n}{d}}\). From the point of view of linear algebra, the structure and duality of the previous objects does not change under such linear embeddings \(i\) and \(j(i)\). However, the embedding \(i\) defines a valuated matroid \(M(X)\) for \(V\), and \(j(i)\) defines a valuated matroid \(M(L(X))\) for \(V^*\), and in general a duality relationship can't be established between these valuated matroids (nor the moduli spaces they represent). The paper studies some instances of the dependency of \(M(L(X))\) on \(i\) and \(j(i)\) and it consists of two parts: the first one contains the main results, while the second one is mostly devoted to worked examples (almost every one of them using the trivial valuation) and connections to other tropicalizations of moduli spaces. The main results include Theorems 8 and 18, and these are presented first for the trivial valuation case. The first one characterizes the matroid \(M(L(X))\) in three different ways, including a geometric characterization using the matroids of lines associated to the hyperplane arrangement \(\mathcal{A}(X)\) on \(X\) induced by the set of coordinate hyperplanes of \(K^n\). The second one is a more specific characterization of what could be called the \textit{generic} matroid \(M(L(X))\), since this matroid is the same for all \(X\) in a Zariski-open set \(G(d+1,K^n)\). Such matroid can be described using a construction called the \textit{Dilworth truncation} applied to the free (uniform) matroid \(U_{n,n}\). In fact, the closed embedding \(j(i)\) induces non-trivial equations on \(G(d+1,K^n)\), and the set of spaces \(X\in G(d+1,K^n)\) for which \(M(L(X))\) is the previous Dilworth truncation is the (possibly) proper subset of elements \(X\in G(d+1,K^n)\) away from the vanishing locus of these new relations. These equations are described in Example 13 for the case \(d+1=3\), \(n=6\). The proof of Theorem 8 relies mostly on linear algebra arguments, and is just about unwinding the correct concepts and definitions, since we have concrete matrices to work with. In fact, the proof is independent of the field being valued, so this result can be adapted to the case in which the valuation \(\text{val}\) is not necessarily trivial by taking the valuation of the Plücker vector of the matroid \(M(L(X))\) so obtained. The proof of Theorem 18 uses a geometric interpretation of the Dilworth truncation for \(\mathbb{C}\)-realizable matroids, together with the description of \(M(L(X))\) as the matroid of lines associated to \(\mathcal{A}(X)\) from Theorem 8. Reviewer's remark: It is assumed (without specifying it) that the closed embedding \(i:V\hookrightarrow{}K^n\) has non-degenerate image, this is, that \(X\) is not contained in any coordinate hyperplane of \(K^n\). In this case, these coordinate hyperplanes induce an arrangement of hyperplanes \(\mathcal{A}(X)\) on \(X\).