Categorical Statistical Mechanics
A GCL research programme about composition, probability, equilibrium and the passage from microscopic systems to macroscopic behaviour.
Programme: CSM-001 · Status: active research, not a certified mathematical result · Authority: Human Steward approval, 4 November 2026 · Tracker: MATH-PROGRAMME #3136.
The question
Statistical mechanics describes large collections of interacting objects by assigning weights or probabilities to microscopic configurations. We ask whether a common language of systems, interfaces, composition and limits can reconstruct these familiar calculations, then reveal new theorems about which physical properties survive a change of scale.
Category theory offers a precise vocabulary for composition. But a categorical diagram cannot, by itself, prove that a thermodynamic limit exists or that a phase transition occurs. The programme is designed to preserve that distinction.
A worked example
Put spins \(s_i\in\{0,1\}\) on a periodic chain, with a statistical weight of \(2\) for each pair of equal neighbours and \(1\) for each unequal pair. Write the transfer matrix
When \(N\) links are composed and the endpoints identified, the partition function is
For example, for three spins, \(Z_3=28\). Matrix multiplication sums over intermediate states while multiplying local weights, exactly as required to evaluate the partition function. The example is established mathematics, used here as an explicit baseline, not a new GCL result.
This uses non-normalized weights. If every component were turned into a stochastic matrix too early, the discarded normalization factors would erase information required for free energies.
The main obstruction: the microscopic theory may not survive coarse-graining
A coarse-graining map \(C:X\to Y\) lumps detailed configurations \(x\) into macrostates \(y\). For finite sets and weights \(w(x)\), the effective weight is exactly
The total weight does not change. Yet the effective interactions can become nonlocal, and infinite-volume coarse-graining can produce measures that are not Gibbsian for any reasonable local interaction. Exact composition does not imply physical locality.
The most important CSM research question is therefore not whether finite marginalization works. It does. The question is which additional mathematical conditions preserve the physical structures of Gibbs equilibrium when systems become large or infinite.
Research lanes and gates
| Gate | Work | Status |
|---|---|---|
| G0 / CSM-00 | Audit categorical probability, thermostatics, Gibbs specifications and renormalization prior art | Open |
| G1 / CSM-01 | Define precise finite semantics; independently replay weighted networks, Ising systems and exact marginalization | Initial executable specimen supplied; full formal semantics open |
| CSM-02 | Relate entropy maximization and Legendre duality to the categorical semantics | Open |
| G2 / CSM-03 | Obtain a new theorem or obstruction for coarse-graining under explicit hypotheses | Open; decisive promotion gate |
| G3 / CSM-04 | Analyze DLR specifications, quasilocality, extremal phases and thermodynamic limits | Open |
| G4 / CSM-05 | Extend selectively to nonequilibrium dynamics and quantum operator-algebraic settings | Deferred until earlier gates justify expansion |
What the literature already establishes
Existing research has laid important foundations. Fritz develops Markov categories for probability. Baez, Lynch and Moeller give compositional thermostatics through entropy-maximizing operadic structures. Sergeant-Perthuis studies categorical Gibbs specifications and phases. Van Enter, Fernández and Sokal establish rigorous non-Gibbsian renormalization pathologies.
These are prior results, not discoveries credited to CSM. The programme's source ledger and technical specification delineate the hypotheses and prospective novelty.
What happens next
The first bounded research obligation is to finish an exact finite model: a typed finite category with explicit tensor and trace conventions, independently checked calculations, examples with fields and different boundary conditions, and a source-pinned proof of the claimed semantics. The initial pure-Python specimen and regression tests provide only a small starting point.
The decisive later obligation is a theorem explaining when coarse-graining preserves Gibbs structure and identifying where it fails, without repackaging established non-Gibbsian examples as original results.
Claim status: Programme authorization and repository admission are administrative/documentary acts. No theorem is asserted to be new, proved by GCL, independently validated or certified.
Programme tracker · Source and novelty register · Technical SPEC