In a chemical reaction network governed by mass-action kinetics, two structures define the system's long-term behavior. Conserved quantities — linear combinations of species concentrations that remain constant along any trajectory — determine what the system preserves. Siphons — subsets of species that, once depleted to zero, stay at zero forever — determine what the system can permanently lose.
The paper conjectures that these two structures are dual. Specifically, siphons and preclusters (subsets defined by conservation laws) are related by a precise topological inversion. The network's preservation structure and its extinction structure are not independent features. They are two descriptions of the same underlying connectivity.
The duality is structural, not quantitative. It doesn't say that what is preserved in one part of the network is lost in another. It says that the capacity for permanent loss (siphons) and the capacity for permanent conservation (conserved quantities) emerge from the same graph-theoretic properties. A network whose topology supports rich conservation laws is, by the same topology, a network with specific vulnerability patterns.
This makes a prediction: learning a network's conserved quantities should immediately reveal its extinction modes, and vice versa. The two analyses are not separate studies of different phenomena. They are the same analysis read in opposite directions.
The through-claim: persistence and vulnerability are not opposing properties to be balanced. They are dual descriptions of a single structure. A network is robust precisely where it is fragile — the conservation law that stabilizes one subspace defines the siphon that threatens another. Knowing where a system cannot change tells you where it can permanently break. The same wiring diagram that guarantees conservation guarantees the possibility of irreversible loss.