The structure of the moduli spaces of toric dynamical systems

The uploaded paper proves that for a weakly reversible graph G the toric locus V(G) is homeomorphic to [(x0+S)∩R^n_{>0}] × B(G), via the explicit map ϕ(x,β) = (β_{y→y′}/x_y)_{y→y′∈E}. It establishes bijectivity (Propositions 5.7 and 5.8) and continuity of both ϕ and ϕ^{-1} using the continuity of Q_{x0} (Theorem 3.5) and the continuity of the auxiliary map q̂, hence concluding ϕ is a homeomorphism (Corollary 5.12) . The continuity of Q_{x0} is supported by a careful three-step argument (using matrix-tree polynomials K_i and a transversality result), see Theorem 3.5 and its proof . The candidate solution gives essentially the same forward/backward maps, but its continuity proof for Ψ relies on an incorrect linear-algebra step: it asserts a fixed direct-sum decomposition R^m = U ⊕ W to project log T(k) − Y^T log x0 onto W. In general, when the deficiency δ > 0, U + W has dimension m − δ < m, so R^m ≠ U ⊕ W and the claimed projection does not exist; consequently, the “fixed linear operator” representation of log x*(k;x0) is unjustified. The paper’s proof avoids this pitfall by using the continuity of Q_{x0} and the continuity of q̂, so the paper is correct while the model’s proof contains a critical flaw.