Published September 2, 2026
| Version v1
Preprint
Open
Topological Tax and the Fluctuation Theorem
Description
The detailed fluctuation theorem (DFT) $p(\sigma)/p(-\sigma)=e^{\sigma}$ constrains the full statistics of entropy production. We show that the geometric and information-theoretic mechanism underlying these constraints is a topological-information structure built from two objects: holonomy and tax.
First, holonomy: the poles of the DFT mean function $m(a)=a\tanh(a/2)$ sit at the odd squares $\pi^2(2k-1)^2$, exactly the Laplacian spectrum of a ring with antiperiodic boundary conditions ($\mathbb{Z}_2$ holonomy $-1$)---so the DFT itself is the Radon--Nikodym form of loop-orientation reversal. Second, the topological tax $\Delta\tau$: the entropy rate of the trajectory's homology record on the abelian cover, quantifying the trajectory's topological information cost, decoupled from the dissipation magnitude.
For finite irreducible Markov jump processes we prove: no topology, no tax ($\Delta\tau>0$ iff $b_1\ge1$); the $\sigma$ coordinate reproduces the DFT moment body (self-contained); the tax coordinate carries a rate-independent coding floor and an integer-quantized capacity; and exact factorization on the occupation lattice, together with factorization of the reachable regions of even/odd observables, which explains why bidirectional estimation is necessary.
Exact factorization of the rate component remains open: it holds at equilibrium and perturbatively, while far from equilibrium a residual magnitude--type coupling appears; in the slow-chord limit the defect does not vanish but is characterized completely in closed form, delineating the boundary of exact factorization.