A counterexample on tropical linear spaces (Q331148)

Tropical geometry has become an important branch of real algebraic geometry over the last decade or so. An important ingredient is the \textit{tropicalization map} taking elements of a field \(K\) (where classical geometry can be defined) to \(\mathbb{R}\) by a non-Archemedean valuation \(v\). Suppose \(S \in K[x_1, \ldots, x_n]\) is a set of polynomials generating an ideal \(I\) in the polynomial ring, then the fundamental theorem of tropical varieties [Speyer-Sturmfels] states that the tropicalization of \(S\) is the intersection of that of the generators \(f\) of \(I\). For an \(m \times n\) matrix \(X\), let \(J_k\) be the set of determinants of all the \(k \times k\) submatrices of \(A\), then the tropical rank of the associated matrix \(C\) is the largest integer \(r\) such that \(C\) is not in the intersection of the tropicalizations of \(f\). If \(S\) is a finite set and the corresponding tropical variety is cut out by \(f\), then \(S\) is called a tropical basis of \(I\). A theorem of Richter-Gebert, Sturmfels and Theobald gives a sufficient condition for \(S\) to be a tropical basis: for an \(m \times n\) matrix \(A\) of rank \(k\) over \(K\) with all submatrices formed by \(k\) columns of \(v(A)\) having tropical rank \(k\). Then, the rows of \(A\) form a tropical basis if \(m=k\). Can condition \(m=k\) be removed? Yu and Yuster conjectured so (cf.~Conjecture 2). In this short note, the author nicely constructs an explicit counter-example to this conjecture.