Abstract

Vassena and Stadler localized instability in reaction networks to unstable cores: minimal child-selection matrices with an eigenvalue of positive real part. For the sequential distributive dual futile cycle they found exactly three unstable cores, all unstable-positive, all with determinant of modulus one, and conjectured that the same picture persists for the nn-site cycle built by adding phosphorylation steps. We settle this conjecture by separating its two assertions. The determinant assertion holds in a class much larger than the futile cycles: every child-selection matrix of an elementary enzyme-conversion system, singular or not, has determinant 00 or ±1\pm1. The structural assertion fails in two ways. For every n3n\ge3 the cycle contains a fixed six-species minimal unstable-negative core, and for every n2n\ge2 it contains a minimal unstable-positive core of dimension 2n+12n+1 that specializes at n=2n=2 to one of the three dual-cycle cores and remains minimally unstable under every positive column scaling; the literal dimensions of minimal positive cores are therefore unbounded. A five-species restriction of the negative witness is strictly stable at unit scaling yet is a DD-unstable-negative core, so ordinary and scaling-invariant minimality differ within one network. For the positive family we derive a scalar return equation that identifies the spectral abscissa for arbitrary rates and losses, giving exact rate comparisons and a sharp loss threshold; at unit rates the unstable eigenvalue is W(n1)/(2(n1))+O(W(n1)2/(n1)2)logn/(2n)W(n-1)/(2(n-1))+O(W(n-1)^2/(n-1)^2)\sim\log n/(2n), so the instability weakens as the cores grow. An explicit saturating-kinetic model of the complete three-step network realizes an unstable positive equilibrium with exactly three eigenvalues in the open right half-plane. The determinant theorem, both core families and the scaling statements are verified in Lean 4; the rate-dependence, asymptotic and kinetic results have conventional proofs with exact rational certificates that a supplied script replays.