The tropical resultant (Q2378643)

Any mathematics worked on the set \(\mathbb{T}:=\mathbb{R}\cup\{-\infty\}\) with addition \(a\oplus b:=\max\{a,b\}\) and multiplication \(a\otimes b:=a+b\) are called \textit{tropical}. In this review, we will use the adjective \textit{classical} as opposed to tropical. Many (but not all!) classical definitions carry over to the tropical setting, just replacing \(+\) by \(\oplus\) and \(\times\) by \(\otimes\). Tropical mathematics turns out to be a coarse shadow of classical mathematics, so that a strategy to attack an extremely difficult problem in classical mathematics is to solve the tropical version of the problem and then \textit{lift} the result from the tropical to the classical world. The paper under review is concerned with the definition and properties of the resultant of two univariate polynomials, in the tropical setting. Classically, if \(f,g\) are polynomials in \(X\) with complex coefficients, the resultant of \(f\) and \(g\) is defined as the determinant of a certain matrix depending on the coefficients of \(f,g\), called the \textit{Sylvester matrix of \(f,g\)}. It is denoted \(R(f,g)\). It is well-known that \[ R(f,g)=a^m b^n\prod_{i=1}^n\prod_{j=1}^m(\alpha_i-\beta_j), \] where \(a,b\) are the leading coefficients of \(f,g\) and \(\alpha_1,\ldots,\alpha_n\) and \(\beta_1,\ldots,\beta_m\in\mathbb{C}\) are the roots of \(f,g\) respectively. The author defines the resultant in the tropical setting as the tropical determinant (or tropical permanent) of the Sylvester matrix and proves an equality which is exactly the tropical counterpart of the equality above. Some remarks are now in order. First, it is explained in the paper that different tropical polynomials (in any number of variables) \(F,G\in \mathbb{T}[X_1,\ldots,X_s]\) can give rise to the same convex piece-wise linear function, i.e., \(F(p)=G(p)\in\mathbb{T}\), for all \(p\in\mathbb{T}^s\). Two such tropical polynomials are called \textit{equivalent} and any equivalence class is called a \textit{tropical polynomial function}. The author actually states and proves his results for tropical polynomial functions. He also proves that the right hand side of the equality is tropically singular (at the coefficients of \(f,g\)) if and only if the left hand side is (at \(\alpha_1,\ldots,\alpha_n,\beta_1,\ldots,\beta_m\)). Tropical singularity means that the maximum is attained, at least, twice. Second, the proofs in this paper are \textit{not} (and cannot be) mere translations of the classical ones, meaning that the paper is not at all trivial. An important ingredient in the paper is the tropical version of the \textit{fundamental theorem of algebra}, which naturally holds for tropical polynomial functions, rather than for polynomials. This is the reason why the coefficients of a generic tropical polynomial \(p\) cannot be written as symmetric tropical polynomial expressions in the roots of \(p\) (Cardan formulas). The proofs in the paper involve an induction step on the degree of the pair \((f,g)\) and a careful up-grading process on the coefficients of \(f,g\) that allow the author to obtain some sort of Cardan type formulas. Reviewer's remark: Going into finer detail, the paper would be easier to read if remarks 1 and 3 and other statements had longer explanations. Moreover, the bibliography is, perhaps, too short; e.g., it does not include important well-established sources, such as the 1992 book by \textit{F. L. Baccelli, G. Cohen, G. J. Olsder} and \textit{J. P. Quadrat} [Synchronization and linearity. An algebra for discrete event systems. Chichester: Wiley (1992; Zbl 0824.93003)], where the tropical fundamental theorem of algebra is treated in p. 123. The notation is confusing because no clear distinction is made between classical and tropical operations. Some authors (who find it tiresome to write \(\oplus\) and \(\otimes\) repeatedly) use quotation marks for tropical statements. In this vein, we could have written above that the author proves \[ \lq\lq R(f,g)"=\lq\lq a^m b^n\prod_{i=1}^n\prod_{j=1}^m(\alpha_i-\beta_j)". \] In my opinion, the notation must indicate whether the operations are classical or tropical. The English used is wrong sometimes (see e.g., \`\` the followings'' in p. 95, column 1, line 35). The paper contains misprints, such as p. 94, column 2, line 8, it should say ``\(R(F,G)\geq a_n^m\oplus b_m^n\)''. All in all, the results are interesting and the paper is worth reading. Another approach to the tropical resultant can be found in \textit{L. F. Tabera}'s paper [Rev. Mat. Iberoam. 24, No. 3, 941--961 (2008; Zbl 1216.14048)].