Autocatalytic Closure for the Working Mathematician

In which emergence turns out to be a theorem.

1. You Already Know This

You have encountered this shape before. Consider three structures you command:

Generated groups. Given elements a, b in a group G, the subgroup ⟨a, b⟩ is the smallest subgroup of G containing a and b. You specify generators; the group structure determines what must exist.

Eigenspaces. Given a linear operator T on V, the eigenspace E_λ = ker(T − λI) is not chosen—it falls out of the operator. You specify the relation (T); the domain (E_λ) emerges.

Algebraic closure. The algebraic closure Q is not given in advance. You specify the property (every polynomial splits); the field that satisfies it is the smallest such extension. The domain is the answer to a closure condition.

In each case, you did not begin by specifying a domain. You specified a relation or property, and the domain fell out as a fixed point. This is routine mathematics—you do it without noticing the pattern.

The left column is comfortable. The right column is exactly the same mathematics—closure under a specified operation—but you may not have noticed how often you already work in it.

2. A Puzzle

Here is a small reaction network. Molecules are circles; reactions are diamonds. Solid arrows connect reactants to reactions and reactions to products. Dashed red arrows show catalysis—which molecule enables which reaction. Molecules in blue are food: freely available from outside.

Six molecules: f₁, f₂ (food), a, b, c, d. Five reactions. One reaction (r₅) needs molecule d as both reactant and catalyst—but nothing produces d. One pair (r₁ and r₂) forms a mutual dependency: r₁ needs b as catalyst; r₂ needs a as catalyst and produces b; but a is produced by r₁.

The question: which reactions can sustain themselves from the food set alone?

3. The Mathematician's Instinct (And Why It Stalls Here)

Your training tells you: define the domain, then prove theorems about it. Fix a group G, then study its structure. Fix a topological space X, then classify its coverings. The domain comes first; results follow.

Here, this instinct stalls. What "exists" in the system depends on what can be produced. What can be produced depends on which reactions can fire. Which reactions can fire depends on what exists to catalyze them. You feel a circularity—and your training says: stop. Circularity means ill-definition.

But it isn't circular. It's a fixed-point equation on ℘(R).

You're looking for R′ ⊆ R such that R′ = Φ(R′), where Φ keeps exactly those reactions whose preconditions are met within the closure of R′ itself. The apparent circularity is the self-referential character of any fixed-point condition. You solve these routinely in other contexts—you just haven't seen this one before.

4. The Algorithm (Greatest Fixed Point)

Rather than searching bottom-up through 2^|R| subsets, start from the top. Begin with all reactions; iteratively remove any reaction whose preconditions fail; repeat until stable. This is a greatest fixed-point computation: a monotone map on (℘(R), ⊇).

Step through it on the puzzle from Section 2.

Observe the structure. The operator Φ(R′) = {r ∈ R′ : ρ(r) ⊆ cl_R′(F) ∧ ∃x ∈ cl_R′(F), (x,r) ∈ C} is monotone on (℘(R), ⊇): if R′ ⊇ R″ then cl_R′(F) ⊇ cl_R″(F), so Φ(R′) ⊇ Φ(R″). Tarski guarantees a unique greatest fixed point. You just computed it.

5. Now the Definition

Read the definition again. No intrinsic property of the elements of X appears anywhere. The definition references only the maps ρ, π, and C—which are relational structure on a graph. Identity of a type x ∈ X is exhaustively determined by its relational position: what it catalyzes, what produces it, what consumes it. Nothing else.

This means the definition is substrate-independent. Any system that instantiates the same graph of reactant/product/catalysis relations is, formally, the same CRS. The labels change; the structure is identical.

6. Emergence, Proved

You do not need to accept "emergence" as a philosophical position. You can prove it as a theorem.

This is the precise algebraic definition of an emergent property: a property P of a set S that holds for S but fails for all proper subsets of S. No philosophy. No hand-waving. A property of the lattice.

The collection of all RAFs of a CRS forms a partially ordered set under ⊆, with the maxRAF at the top. The irrRAFs are the minimal non-empty elements. Each irrRAF is a node where emergence first appears—the smallest subsets at which a collective property crystallizes.

The Phase Transition

Kauffman (1986) conjectured, and Mossel & Steel (2005) proved for certain model classes, that in a random CRS with catalysis probability p, the probability that a RAF exists undergoes a sharp transition near p ∼ 1/n:

Below the threshold: no subset of R is self-sustaining. Above: irreducible self-sustenance appears. The transition sharpens with n. This is the moment emergence crystallizes—not as metaphor, but as a phase transition in the lattice of subsets of R.

7. Coda

You didn't need to define the domain first. You needed to specify the relations—the catalysis graph, the reactant/product maps, the food set. The domain fell out as a fixed point of a monotone operator on a finite powerset lattice. Tarski did the rest.

This is how emergence works—not as philosophy, but as algebra. An irreducible RAF is a set whose defining property provably fails for every proper subset. That is a theorem, not an argument. And because the definition is purely relational—no intrinsic properties of X appear anywhere—the theorem holds for any substrate that instantiates the graph. Chemistry, ecology, knowledge, governance: same lattice, same fixed point, same emergence.