OctoInformBeta

OpenAI preprint, September 2026 · §3. From affine labels to persistent trajectories

Lemma 3.1: Directed returns and quantitative weight ordering under a bound on reaction rates

An inward flux criterion for mass action.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Consider the sign of the instantaneous change of a linear concentration level in a mass-action system.

A reaction network NN on dd species consists of a finite complex set C⊂ℕdC\subset\mathbb N^d and a finite reaction set R⊂C×CR\subset C\times C, with distinct source and target in every reaction.✦ A network reaction e=y→ze=y\to z is an ordered pair (y,z)∈R(y,z)\in R.✦ The network NN is weakly reversible if every reaction y→zy\to z has a directed path of one or more reactions from zz to yy.✦

A positive state x∈ℝdx\in\mathbb R^d is a state with xi>0x_i>0 for every species ii.✦ The mass-action monomial xyx^y of a real state xx at a complex yy is the product in (1): xy=∏i=1dxiyi.x^y=\prod_{i=1}^d x_i^{y_i}.(1) The dot product r⋅xr\cdot x of real vectors r,x∈ℝdr,x\in\mathbb R^d is the coordinate sum in (2): r⋅x=∑i=1drixi.r\cdot x=\sum_{i=1}^d r_i x_i.(2) For constant reaction rates κe\kappa_e, the mass-action vector field ff of NN is the function in (3): f(x)=∑y→z∈Rκy→zxy(z−y).f(x)=\sum_{y\to z\in R}\kappa_{y\to z}x^y(z-y).(3)✦ For a slope r∈ℝdr\in\mathbb R^d, the instantaneous change of the linear level r⋅xr\cdot x along ẋ=f(x)\dot x=f(x) is r⋅f(x)r\cdot f(x).

For the graph formulation, let VV be a vertex set and EE a finite edge set, with source and target maps s,t:E→Vs,t:E\to V. A directed return path for an edge ee in the directed graph with vertex set VV and edge set EE is a path of one or more edges from t(e)t(e) to s(e)s(e). Given a level function a:V→ℝa:V\to\mathbb R, vertex weights wv>0w_v>0, and edge rates κe>0\kappa_e>0, the weighted level flux Φ\Phi is the number in (4): Φ=∑e∈Eκews(e)(a(t(e))−a(s(e))).\Phi=\sum_{e\in E}\kappa_e w_{s(e)} \bigl(a(t(e))-a(s(e))\bigr).(4) For a parameter δ≥0\delta\geq0, the rate bound is (5): δ∑f∈Eκf<κefor every e∈E.\delta\sum_{f\in E}\kappa_f<\kappa_e \qquad\text{for every }e\in E.(5)

The Claim

Theorem 1. Let V,E,s,t,a,w,κV,E,s,t,a,w,\kappa be as above, and let δ≥0\delta\geq0. Suppose every edge has a directed return path and the rate bound (5) holds. Suppose also that wv≤δwuw_v\leq\delta w_u whenever u,v∈Vu,v\in V satisfy a(u)<a(v)a(u)<a(v). Then Φ≥0.\Phi\geq0.(6)✦

Alternatively, suppose every edge has a directed return path, the rate bound (5) holds, and ws(f)≤δws(e)w_{s(f)}\leq\delta w_{s(e)} whenever e,f∈Ee,f\in E satisfy a(s(e))<a(s(f))a(s(e))<a(s(f)). Then Φ≥0.\Phi\geq0.(7)✦

Let NN be a weakly reversible reaction network, xx a positive state, and r∈ℝdr\in\mathbb R^d a slope. Let the constant reaction rates κe\kappa_e be positive and δ≥0\delta\geq0. Suppose the rate bound (5) holds with E=RE=R, and xz≤δxyx^z\leq\delta x^y whenever y,z∈Cy,z\in C satisfy r⋅y<r⋅zr\cdot y<r\cdot z. Then r⋅f(x)≥0.r\cdot f(x)\geq0.(8)✦

In the paper, the result reads: let V,EV,E be finite vertex and edge sets, s,t:E→Vs,t:E\to V source and target maps, a:V→ℝa:V\to\mathbb R a level, wvw_v positive weights, κe\kappa_e positive rates, and δ>0\delta>0; suppose every edge has a directed return path, wv≤δwuw_v\leq\delta w_u whenever an ordered vertex pair satisfies a(u)<a(v)a(u)<a(v), and δ∑e∈Eκe<min⁡e∈Eκe\delta\sum_{e\in E}\kappa_e<\min_{e\in E}\kappa_e when E≠⌀E\ne\varnothing; then ∑e∈Eκews(e)(a(t(e))−a(s(e)))≥0\sum_{e\in E}\kappa_e w_{s(e)}(a(t(e))-a(s(e)))\geq0.

What Is Checked

The flux statements (6) and (7) are stated over arbitrary vertex sets with finitely many edges and the specified return paths, weight ordering, and rate bounds, while (8) is stated over finite weakly reversible reaction networks with positive states and the specified monomial and rate bounds, including empty edge and reaction sets.