Skip to main content
Back to timeline
arXivSource publication:

A complete machine-checked proof settles global convergence for detailed balanced mass-action systems: each positive class has one equilibrium and every positive trajectory reaches it

Synopsis

For finite detailed balanced mass-action systems with positive rate constants, a complete chain from persistence to convergence establishes that each positive stoichiometric compatibility class contains exactly one positive equilibrium and that every positive trajectory converges to it. The chain uses a logarithmic barrier for weakly reversible systems as its first step, writes positive equilibria as an exponential family via the Horn–Jackson dissipation identity to obtain uniqueness, and replaces LaSalle's invariance principle with a bounded-speed descent, with every step checked by the Lean 4 kernel.

Source-provided article image: Global Convergence of Detailed Balanced Mass Action Systems
arXiv

Interpretation

In a finite detailed balanced mass-action system with positive rate constants, each positive stoichiometric compatibility class contains exactly one positive equilibrium, and every positive trajectory starting in that class converges to it. Existence, uniqueness and local asymptotic stability of equilibria in such systems were classical, while the global attractor conjecture asked whether every positive trajectory converges to the positive equilibrium in its compatibility class; an existing proof took the route of toric differential inclusions. The argument joins persistence and the entropy argument into one chain, the main theorem is formalized in Lean 4 with Mathlib and quantifies over all global forward solutions, and the three-species example A+B⇌C and 2A⇌B converges to (1/2,1/2,1/2) in the class xA+2xB+3xC=3.

The argument first builds a barrier as a finite minimum of affine functions, whose superlevel sets confine all solutions from a given initial state to a compact box inside the positive orthant. This step reuses the barrier argument for boundedness and persistence of weakly reversible systems, which requires only weak reversibility and positive rate constants, not detailed balance; the detailed-balance convergence argument uses it only through this common compact container. The barrier directions are fixed before h is chosen, so the sign gap is independent of h; this yields one logarithmic bound for all solutions and all coordinates, plus a speed bound on the compact container formed by intersecting the compatibility class with the box.

Positive equilibria are written as an exponential family parameterized by the orthogonal complement of the stoichiometric subspace, which makes them unique within each class. This step comes from the Horn–Jackson dissipation identity: an entropy-type function is non-increasing along positive solutions, and zero dissipation exactly characterizes positive equilibria. Every summand of the dissipation identity is nonpositive and vanishes only when the logarithmic gap across a reaction is zero; the exponential map is bijective on that parameter set, and the difference of two equilibria in one class is orthogonal to the parameter difference, so termwise nonnegativity forces the parameters to agree.

The convergence step replaces LaSalle's invariance principle with a bounded-speed descent argument that places every cluster point in the set of positive equilibria. The LaSalle route needs the omega-limit set to be forward invariant and therefore relies on continuous dependence on initial data; the descent argument here uses only a speed bound. The argument assumes neither uniqueness of solutions nor continuous dependence on initial data; the trajectory stays in the compact container, and if a cluster point had negative dissipation, the entropy would drop below its infimum over a sufficiently short time interval, a contradiction.

Perspective

The conclusions concern finite detailed balanced mass-action systems with positive rate constants and strictly positive initial states; the formalization covers the numbered steps from persistence to convergence and the three-species example, while the asymptotic-stability corollary within a class, the Wegscheider criterion, several discussions and the numerical computations are not formalized.

The persistence step takes as input the barrier argument for boundedness and persistence of weakly reversible systems, whose Lean modules are adapted into this project; in the formalization the compact container is written as the intersection of [ε, ε⁻¹]d with the compatibility class rather than [e⁻L, eL]d with the class, with the same conclusions.