A reaction network in which every reaction lies on a loop never loses a species
Weakly reversible mass-action systems stay between fixed positive bounds for every choice of rate constants.
The result in plain terms
Models in chemistry, cell biology and epidemiology are often written as reactions: species combine and convert at rates proportional to the product of their concentrations. That is mass action, and it turns a reaction diagram into a system of differential equations. The rate constants are rarely known. A modeller wants to know what the diagram alone guarantees.
Two things can go wrong in such a model. A species can run down towards zero, or a concentration can grow without limit. Checking by simulation means trying one set of rates and one starting point at a time.
Write each side of each reaction as a node, such as A + B or 2C, and each reaction as an arrow between nodes. The network is weakly reversible when every arrow lies on a closed loop of arrows. A + B → C → A + B qualifies, and so does A → B → C → A. The theorem says that for a weakly reversible network, with any positive rate constants and any positive starting concentrations, every concentration eventually lies between a positive floor and a finite ceiling and stays there. The floor and ceiling depend on the network, the rates and the conserved totals. They do not depend on where the system started. Nothing dies out and nothing blows up.
What was known
The structural theory began with Horn, Jackson and Feinberg in 1972, who showed that networks satisfying a balance condition on the rates settle to a unique equilibrium. Weak reversibility is a condition on the diagram alone, and whether it prevents extinction was left open. Craciun, Nazarov and Pantea proved the bounded-and-persistent conclusion for two species in 2013. Gopalkrishnan, Miller and Shiu proved it in 2014 for networks whose diagram is a single connected piece, and Boros and Hofbauer gave another proof of that case in 2020. Networks with several pieces and more than two species were not covered.
Why it is usable today
The hypothesis is checked on the diagram. Finding whether every arrow lies on a loop is a standard graph computation that takes time proportional to the size of the network, and for a small model it is done by eye. No rate constant has to be measured and no simulation run.
- Screening models. A weakly reversible model cannot lose a species or diverge under any parameter fit, so those failure modes need not be searched for.
- Designing circuits. A synthetic reaction network built to be weakly reversible carries the guarantee whatever rates the laboratory achieves.
- Reading failures. The Lotka–Volterra predator and prey reactions X → 2X, X + Y → 2Y, Y → ∅ are not weakly reversible, and their populations cycle as close to zero as the starting point dictates.
What it does not give
Many working models contain one-way steps, such as degradation with no matching production, and fall outside the hypothesis. The theorem is for fixed rate constants and for the differential equation. In a cell with a handful of molecules, the count of a species can still hit zero for a time. Staying in a bounded region is weaker than settling down: a weakly reversible network can oscillate for ever. The size of the floor and ceiling is not given by a formula.
Scores and sources
Applied score 5 of 5, difficulty score 4 of 5. The problem dates from about 1974. OpenAI's Lean formalization covers supporting statements of this family, not the main theorem. The scores take the theorem as stated.