RAF Sets: Catalytic Closure Made Decidable

The autocatalytic set made rigorous and checkable — and the one thing it still does not build

Kauffman’s autocatalytic set — treated on the companion page as the origin-of-life threshold — began as an informal proposal, and its original random-catalysis argument was later challenged for appearing to need implausibly much catalysis. RAF theory — Reflexively Autocatalytic and Food-generated sets, formalised by Hordijk and Steel — is what made the idea precise: an exact definition of catalytic closure, an efficient algorithm that detects it in a real reaction network, and instances found in actual chemistry. It is the closest the established literature comes to a decidable test for the thing this framework calls recursion lock. The framework agrees almost entirely — a RAF is recursion lock made checkable — and draws a single line, the same one it draws for the bare autocatalytic set: a RAF is catalytic closure with no self-produced boundary. It is the rigorous form of ignited, but not yet enclosed.

What RAF makes precise

Hordijk, W., & Steel, M. (2004). Detecting autocatalytic, self-sustaining sets in chemical reaction systems. Journal of Theoretical Biology, 227(4), 451–461. Building on Kauffman, S. A. (1971/1993) and extended in later reviews (Hordijk & Steel; Hordijk, Hein & Steel) and applications (e.g. Vasas et al., “Evolution before genes,” 2012). [verify all author names, titles, venues, dates, volume/issue/pagination and DOIs against the primary sources before publication]

A RAF set lives inside a catalytic reaction system: a collection of possible reactions, a set of catalysis relations (which molecules speed which reactions), and a food set F of simple molecules assumed freely available from the environment. A subset of reactions R is a RAF when it meets two conditions at once. [verify the definitions below against Hordijk & Steel]

RAReflexively Autocatalytic. Every reaction in R is catalysed by at least one molecule that is either in the food set or produced by R itself. The catalysis closes on the set: nothing outside has to reach in to make the reactions go.

FFood-generated. Every reactant any reaction in R consumes can be built up from the food set using only reactions already in R. The set can bootstrap all its own inputs from raw supply.

A set that is both is self-sustaining given the food: it manufactures its own catalysts and assembles its own reactants from what the environment provides. Two properties lift this above Kauffman’s informal version, and both matter here:

It is decidable. There is an efficient (polynomial-time) algorithm that finds the maximal RAF in a reaction network. So “does this chemistry contain a self-sustaining, catalytically-closed subset?” is a question one can actually answer for a given network, rather than merely assert. That is the decisive advance: closure stops being a figure of speech and becomes a computable property. [verify the polynomial-time RAF algorithm and the uniqueness of the maximal RAF]

It is plausible and real. Kauffman’s original model appeared to require each molecule to catalyse an implausibly large number of reactions; RAF analysis showed self-sustaining sets emerge when each molecule catalyses only a small number, and RAF structure has since been identified in real chemical and metabolic networks. The idea survived the move from suggestive to testable. [verify the catalysis-rate plausibility result and each claimed empirical instance against primary literature]

Where the framework agrees

A RAF is recursion lock made decidable. A reaction set whose every member is catalysed from within is exactly the closed, self-producing loop the framework calls recursion lock; RAF’s reflexively autocatalytic condition is that loop stated precisely, and its food-generated condition is the framework’s insistence that nothing persists closed — the set runs on throughput drawn from a supply. It reads onto the same bootstrapping ladder as the informal set: catalytic closure is AC₀, mutual catalysis among members is AC₃, and the interlock of loops into a self-sustaining network is AC₄ — a basin of persistence in the framework’s exact sense. Where the framework must assert that a network has ignited, RAF offers a criterion that could, in principle, check it.

RAF even supports the framework’s sharpest claim about the informal set — that the crossing into self-maintenance is a threshold, not a slope. The emergence of a RAF as catalysis increases is a comparatively sharp, percolation-like transition: below a level, no self-sustaining set exists; above it, one appears. That is the discreteness the companion page argues for, arriving from independent mathematics rather than from the framework’s own commitments — which is exactly the kind of ally a structural claim wants. [verify the sharpness / percolation-like characterisation of RAF emergence]

On the structure, the ladder, and the reading of catalytic closure as a real, detectable property, the framework and RAF are in near-complete agreement. RAF is the strongest formal neighbour the framework has for the claim that recursion can genuinely lock — and the only one that can be run as an algorithm on a candidate chemistry.

Where the framework draws its line

The line is not a disagreement with anything RAF proves. It is the layer RAF’s definition stops beneath. Three things a RAF does not have:

1. No boundary — closure without enclosure. A RAF is catalytically closed but open to its surroundings: it draws freely on the food set and has no self-produced membrane dividing it from its medium. That is precisely Condition 3 — the self-made, identity-bearing boundary — left unmet. In the framework’s terms a RAF is the bare autocatalytic set, not the membrane-enclosed protocell: it has ignited (recursion has locked) without yet enclosing itself (the AC₁₁ boundary pivot). A RAF is therefore necessary but not sufficient for sovereignty — the rigorous form of “an attractor, not yet a sovereign.”

2. Structural existence, not dynamical persistence. RAF asks a static, graph-theoretic question — does a self-sustaining closed subset exist in this network, given the food set? The framework’s question is dynamical — what does it cost to keep occupying that set across the sequence, and what happens at the moment the cost cannot be met? RAF carries no maintenance fee (T₂), no stored record of the set’s own recursion (recurcline), and no defined loss event. It certifies that the set can sustain itself given food; it does not price the persistence or model the death. The framework adds exactly that accounting layer on top of the closure RAF certifies.

3. The threshold is about existence, not internal grain. RAF supports “threshold, not slope” for the first question — is there a self-sustaining closed set at all — but honesty requires the qualifier RAF itself supplies: a maximal RAF can contain a nested hierarchy of smaller, irreducible sub-RAFs. So the crossing into having a RAF is sharp, while what sits inside one can be graded. The framework’s discreteness claim is about the existence of the attractor — occupied or not — and must not be overread as a claim that a self-sustaining network has no internal structure. On this point RAF keeps the framework precise rather than contradicting it. [verify the RAF-hierarchy / irreducible-sub-RAF characterisation]

What the pairing buys

The companion page argues that the origin of life is a discrete ignition, not a gradual slope — but leaves open how such a crossing could ever be detected. RAF is the missing instrument: a decidable criterion that, applied to a candidate chemistry, returns whether a self-sustaining closed set is present. It does not supply the framework’s boundary or its ledger — but it makes the framework’s first threshold, ignition, operational in a way informal autocatalysis never could.

Informal autocatalysis says the loop can close. RAF says whether, for this chemistry, it has. The framework says what it then costs to stay closed, and what is lost when the cost goes unpaid. Three layers, one phenomenon — and RAF is the one that can be computed.

A noted direction · offered, not asserted Where a RAF would cross from attractor to sovereign. RAF certifies closure; the framework’s second threshold is boundary self-production. The natural extension, entirely inside RAF’s own apparatus, is a bounded RAF: one whose access to the food set is gated by a compartment the set itself produces and maintains, so that the food-generated condition runs through a self-made boundary rather than an open medium. A RAF that produced and repaired its own enclosing boundary would be the point at which catalytic closure (AC₀) meets the boundary pivot (AC₁₁) — Kauffman’s protocell stated in RAF’s vocabulary. This introduces nothing to the canon and asserts no chemistry; it only names where, on the framework’s reading, RAF would stop being an attractor’s criterion and start being a sovereign’s. It is the same gap this site’s own Eco 2.0 harness runs at: the maintenance bill met, the self-produced boundary still the open problem.

Where this sits, and what it leaves open. This page is the rigorous complement to the informal origin-of-life entry: Autocatalytic Sets and the Origin of Life reads Kauffman’s threshold; this page reads the decidable criterion that can, in principle, locate it. Together with random Boolean networks they make up the framework’s treatment of the Kauffman lineage — closure, its detection, and the free cycle of a fixed map, read as three different questions about the same body of work. Deliberately left open:

The chemistry of the crossing — what actually tips a network into possessing a RAF — is the live scientific problem, and the framework describes the shape of the crossing without claiming its mechanism.
The bounded-RAF direction above — closure plus a self-produced boundary — is noted as a direction, not a result, and belongs to any future treatment of where RAF meets Condition 3.

None of this touches the framework’s fixed foundations; the six substrate conditions are closed and are not at issue on this page.

Adjacent-work assessments state where Principia Attractum agrees with and departs from neighbouring frameworks. They introduce no constructs and modify no canon; they locate the framework relative to its field, and — where a neighbour’s apparatus invites it — may note a proposed direction, marked as offered rather than asserted. The reading of a RAF as recursion lock made decidable, and as an attractor’s criterion short of sovereignty, is a downstream application of the framework’s existing distinctions, not an addition to them. Every mathematical, chemical and historical claim about RAF on this page, and every citation, is marked [verify] and must be confirmed against the primary sources before publication.

← Back to Relation to Adjacent Work