Tibor Gánti (1933–2009), a Hungarian biochemist, spent his career asking what may be the sharpest question in origins-of-life research: what is the minimal chemical system that constitutes life? Not what molecules life happens to use—not DNA, not proteins, not lipids specifically—but what organization of chemistry makes a chemical system alive rather than merely reactive.
He published his answer in Hungarian in 1971. The book, Az élet principíuma (The Principle of Life), was virtually unknown outside Hungary for three decades. An English translation appeared only in 2003. By then, the origins-of-life field had fractured into warring camps—metabolism-first, replicator-first, membrane-first—each promoting one subsystem as the key innovation. Gánti had already shown, quietly, in a language few Western scientists read, that the question was malformed. You cannot have any one without all three.
His model, the chemoton, is a concrete chemical system—not an abstraction like Rosen’s (M,R) systems, not a definition like autopoiesis, not a graph-theoretic property like RAF closure. It is a specific, simulatable architecture consisting of exactly three coupled subsystems, each necessary, none sufficient alone, locked together by stoichiometry into a unit that metabolizes, replicates, and divides.
The chemoton consists of three autocatalytic subsystems. Each is, in isolation, a known kind of chemistry. The innovation is their coupling.
Each subsystem is chemically unremarkable on its own. Autocatalytic cycles exist in prebiotic chemistry. Template-directed polymerization has been demonstrated in the lab. Lipid vesicles form spontaneously. The question is not whether each can exist, but whether all three can run simultaneously in a single compartment, coupled by nothing more than shared stoichiometry.
This is Gánti’s central innovation, and it is more subtle than it first appears. The three subsystems are not connected by regulatory signals, feedback loops, or allosteric enzymes. They are connected by stoichiometry—by the simple fact that one subsystem’s products are another’s substrates, in exact quantitative ratios.
The coupling works as follows. The metabolic cycle runs as long as food X is available, producing V′ and T as byproducts. V′ is consumed by template replication: the template pVn acts as a sink for V′, pulling the metabolic cycle forward by depleting its product. This is Le Chatelier’s principle doing the work of regulation—no gene expression, no signaling cascades, just mass action. The template’s length determines how much V′ it consumes, and therefore how many metabolic cycles must turn before replication is complete.
Meanwhile, T molecules accumulate in the membrane. As surface area grows while volume grows more slowly, osmotic pressure builds inside the vesicle. When the membrane has incorporated enough T molecules to double its surface area—and this happens at exactly the time when the template has finished replicating, because both consume products from the same metabolic source at fixed ratios—the vesicle becomes mechanically unstable and divides.
Division is not programmed. It emerges from physics: a sphere whose surface area has doubled while its volume has less than doubled must either bud, elongate, or split. The daughter cells each inherit one copy of the template, roughly half the metabolic intermediates, and half the membrane. The cycle begins again.
The simulation below shows the three subsystems running in a single compartment. Food X enters through the membrane; waste Y exits. The metabolic cycle (gold) produces V′ monomers (cyan dots) that are consumed by the template, and T molecules (green) that incorporate into the membrane. Watch all three subsystems grow in lockstep, then divide.
Gánti’s deepest conceptual contribution was not the chemoton itself but the category he invented to describe it: the fluid automaton. Classical automata—clocks, Turing machines, digital computers—are built from persistent components. Gears do not dissolve between ticks. Transistors are not rebuilt every clock cycle. The hardware endures while the software runs on it.
A fluid automaton has no persistent components. Every molecule in the chemoton is continuously created and destroyed. The metabolic intermediates A1 through An are consumed as fast as they are produced. The template pVn is a transient structure, built from monomers that were themselves recently synthesized. Even the membrane is in constant flux—T molecules exchange with the interior, are degraded, replaced. Nothing persists. And yet the organization persists: the pattern of relationships among the three subsystems, the stoichiometric ratios, the cycle of growth and division.
This is Gánti’s version of the ship of Theseus, but sharper: not just “are the planks the same?” but “there are no planks.” The ship is a standing wave in a river of timber. The chemoton is a standing wave in a river of chemistry.
This leads to a profound philosophical disagreement with Rosen that illuminates both thinkers. Gánti says organisms are machines—fluid machines, but machines nonetheless. The chemoton is fully describable by ordinary differential equations. It is simulatable. It is a mechanism, just a peculiar one whose parts are processes rather than objects. Rosen says organisms are not machines—they are closed to efficient causation in a way that no mechanism can be. The same phenomenon—organizational closure, self-production, the persistence of pattern through the flux of matter—leads to opposite philosophical conclusions. Gánti sees a new kind of machine; Rosen sees the end of mechanism.
They are both right about different things. Gánti is right that the chemoton is simulatable—you can write the ODEs, run the numerics, watch the division. Rosen is right that something about the organization resists decomposition—you cannot understand the chemoton by studying any subsystem in isolation. The whole is not the sum. The disagreement is not empirical but metaphysical: what counts as a “machine”?
Click any cell for an expanded explanation. Four frameworks for minimal life, each capturing a different face of the same phenomenon.
| Chemoton | (M,R) Systems | RAF Sets | Autopoiesis | |
|---|---|---|---|---|
| Creator | Gánti 1971 | Rosen 1958 | Kauffman / Hordijk–Steel | Maturana & Varela 1972 |
| Abstraction | Concrete chemistry | Category theory | Set / graph theory | Organizational definition |
| Closure type | Stoichiometric | Efficient causation | Catalytic | Self-production |
| Boundary | Explicit (membrane) | Implicit | None | Required (definitional) |
| Heredity | Template pVn | Not addressed | Not addressed | Not addressed |
| Simulatable? | Yes (ODE, stochastic) | Controversial | Yes (polynomial) | Too abstract (usually) |
| Organism = machine? | Yes (fluid automaton) | No (impredicative) | N/A | No (organizational) |
Manfred Eigen showed in the 1970s that error-prone replication sets a hard limit on template length. Every replication event introduces errors. If the error rate per base per replication exceeds 1/n (where n is the template length), the information in the template dissolves into noise—the error catastrophe. This creates a paradox: long templates need accurate replication, but accurate replication requires complex enzymes, which require long templates to encode them.
Eigen’s own solution was the hypercycle: a system of replicators where each one catalyzes the replication of the next, forming a cooperative loop. But hypercycles are vulnerable to parasites—short-cutting replicators that receive catalytic support without contributing. Without spatial structure, a single parasite destroys the whole hypercycle.
The chemoton offers a different solution. Zachar & Szathmáry (2010) and Zachar et al. (2011) showed that stoichiometric coupling within a compartment allows competing templates to coexist. The membrane creates the spatial structure that prevents parasites from spreading between lineages. Templates compete within a chemoton (intragenomic conflict) but are selected between chemotons (group selection). Compartmentalization does the work that hypercyclic cooperation was supposed to do, but more robustly.
This is one of the chemoton’s strongest claims to biological relevance: it provides a concrete mechanism for the origin of compartmentalized genomes, the step that Szathmáry and Maynard Smith identified as the first major transition in evolution.
Gánti’s answer to the “metabolism first vs. replicator first vs. membrane first” debate is that the question is malformed. Each subsystem alone fails in a specific and instructive way.
Disable any one subsystem in the simulator above and watch the system collapse. This is not a design choice—it is a stoichiometric necessity. The three subsystems are not independent modules bolted together; they are a single chemical process that happens to have three distinguishable aspects.
The chemoton is a model of minimal life, not actual life. The gap between the chemoton and the simplest modern cell is enormous, and understanding that gap is understanding what evolution had to invent.
Jan-Hendrik Hofmeyr (2021) showed that Gánti’s three subsystems, Rosen’s three mappings (f, Φ, β), and von Neumann’s self-reproducing automaton all converge on the same three-part architecture. The metabolic cycle maps to Rosen’s metabolism f (converting inputs to outputs). The template maps to Rosen’s β (the component that closes the loop by carrying the instructions for self-reconstruction). The membrane maps to the organizational boundary that distinguishes self from non-self. Three independent thinkers, working in different decades with different formalisms, arrived at the same tripartite structure.
This convergence suggests that three-subsystem closure is not an accident of carbon chemistry but a necessary architecture for any system that metabolizes, reproduces, and maintains an identity. If artificial life is possible on other substrates—in silicon, in virtual chemistry, in alien biochemistry—it will need some version of all three.
This is Part III. Here is what each part of the arc has established, and how they fit together.
Part I: Autocatalytic Sets showed that catalytic closure—every reaction catalyzed by something in the system—emerges easily in random chemical networks once catalytic probability exceeds a phase transition. RAFs are the mathematical scaffold: self-sustaining sets can bootstrap from a food set without external enzymes. But RAFs have no boundary, no heredity, and no mechanism for producing their own catalysts (only for having them present).
Part II: (M,R) Systems showed that organizational closure goes deeper than catalytic closure. Rosen’s closure to efficient causation requires that every function be produced from within—not just present, but made. This is a mathematical proof that living organization is self-referential in a specific way: the system produces the processors that process the inputs that produce the processors. But (M,R) systems are abstract—they tell you what closure means without showing what it looks like in chemistry.
The chemoton realizes both closures in concrete chemistry. Its metabolic cycle is an autocatalytic set (RAF). Its three-subsystem architecture achieves closure to efficient causation ((M,R)). Its membrane achieves the self-produced boundary demanded by autopoiesis. Its template provides the hereditary information that none of the other frameworks address. Assembly theory would measure the complexity of the chemoton’s products. Persistent homology would detect the topological features of its reaction network. Each framework captures a different face of the same phenomenon: organizational closure—the pattern that persists through the flux of matter, the process that produces the conditions for its own continuation.
Autocatalytic Sets — RAF theory, the bootstrap problem, and self-sustaining chemical networks
(M,R) Systems — closure to efficient causation, metabolism-repair, and why organisms are not machines
Five Levels of Self-Production — from complexity to agency
Before Replication — what came before Darwinian evolution
Function vs Process — the distinction Rosen built on