[Paper Review] Sign conditions for injectivity of generalized polynomial maps with applications to chemical reaction networks and real algebraic geometry

Stefan C. Müller, Elisenda Feliu+4 more|arXiv (Cornell University)|Nov 21, 2013

Advanced Differential Equations and Dynamical Systems55 references7 citations

TL;DR

This paper establishes necessary and sufficient sign vector conditions for injectivity of generalized polynomial maps with real exponents, enabling the preclusion or guarantee of multiple positive steady states in chemical reaction networks and extending Descartes' rule of signs to multivariate systems. The key contribution is a sign-based criterion that generalizes classical injectivity conditions and connects to oriented matroid theory and real algebraic geometry.

ABSTRACT

We give necessary and sufficient conditions in terms of sign vectors for the injectivity of families of polynomial maps with arbitrary real exponents defined on the positive orthant. Our work relates and extends existing injectivity conditions expressed in terms of Jacobian matrices and determinants. In the context of chemical reaction networks with power-law kinetics, our results can be used to preclude as well as to guarantee multiple positive steady states. In the context of real algebraic geometry,our work recognizes a prior result of Craciun, Garcia-Puente, and Sottile, together with work of two of the authors, as the first partial multivariate generalization of the classical Descartes' rule, which bounds the number of positive real roots of a univariate real polynomial in terms of the number of sign variations of its coefficients.

Motivation & Objective

• To develop sign vector-based criteria for injectivity of generalized polynomial maps with arbitrary real exponents on the positive orthant.
• To apply these criteria to preclude or guarantee multiple positive steady states in chemical reaction networks with power-law kinetics.
• To extend classical Descartes' rule of signs to the multivariate setting via sign vectors and oriented matroids.
• To provide algorithmic tools for checking injectivity using sign vectors, chirotopes, and linear programming techniques.
• To unify and generalize existing injectivity conditions based on Jacobian matrices and determinants in the context of parametrized polynomial maps.

Proposed method

• Define generalized polynomial maps $ f_{ ho}(x) = A_{ ho} x^B $, where $ A \in \mathbb{R}^{m \times r} $, $ B \in \mathbb{R}^{r \times n} $, and $ \rho \in \mathbb{R}_+^r $, with real exponents.
• Introduce injectivity with respect to a subspace $ S \subseteq \mathbb{R}^n $, ensuring no two distinct points in a coset $ x' + S $ map to the same value.
• Use sign vectors $ \sigma(S) $ of subspaces to characterize injectivity, particularly $ \sigma(\ker A) \cap \sigma(\operatorname{im} B) = \{0\} $ for unrestricted injectivity.
• Apply oriented matroid theory to compute and analyze sign vectors, especially covectors and cocircuits, using chirotopes and software like Topcom.
• Reduce injectivity testing to feasibility of systems of strict linear inequalities, using exact linear programming and Farkas’ lemma for rational certificates.
• Leverage existing computational tools such as QSopt_ex, RAGLib, and mixed-integer programming to enumerate sign vectors and test for common nontrivial sign vectors.

Experimental results

Research questions

• RQ1 What sign vector conditions ensure injectivity of generalized polynomial maps with real exponents on the positive orthant?
• RQ2 How can sign vector conditions be used to preclude or guarantee multiple positive steady states in chemical reaction networks with power-law kinetics?
• RQ3 In what way does this work generalize Descartes' rule of signs to the multivariate case?
• RQ4 What is the relationship between the sign vectors of $ \ker A $ and $ \operatorname{im} B $ in determining injectivity of the map $ f_{\rho} $?
• RQ5 Can the injectivity of generalized polynomial maps be algorithmically determined using sign vectors and oriented matroid theory?

Key findings

• The paper establishes that injectivity of $ f_{\rho} $ with respect to a subspace $ S $ is equivalent to the absence of a common nontrivial sign vector in $ \sigma(\ker A) \cap \sigma(\operatorname{im} B) $.
• For unrestricted injectivity, the condition $ \sigma(\ker A) \cap \sigma(\operatorname{im} B) = \{0\} $ is both necessary and sufficient when $ A $ and $ B $ have full row and column rank, respectively.
• The work generalizes Descartes' rule of signs to multivariate systems by recognizing sign vector conditions as a partial multivariate extension of the classical univariate rule.
• The injectivity problem reduces to checking the infeasibility of a system of strict linear inequalities, which can be addressed via exact linear programming and Theorems of the Alternative.
• The number of sign vectors can grow exponentially with dimension, but efficient algorithms exist for computing cocircuits and generating all covectors from them.
• Software tools such as Topcom, RAGLib, and QSopt_ex enable practical implementation of the sign vector approach, particularly for testing existence and uniqueness of positive solutions in reaction networks.