Local thermodynamic equilibrium is the assumption that makes non-equilibrium thermodynamics tractable. A system in a non-equilibrium steady state — say, a metal bar with one end heated and the other cooled — is not in global equilibrium. But each small piece of the bar, the assumption says, is locally in equilibrium: it has a well-defined local temperature, and the thermodynamic relations (equations of state, phase boundaries, free energies) hold at that local temperature. This lets you describe the non-equilibrium system as a patchwork of equilibrium pieces.
Note (2026-11-14; read against the paper's full text): the essay presents a detailed-balance mechanism and says reversing the current shifts the interface temperature the other way. The paper is a numerical study of a one-dimensional Hamiltonian-Potts model with fractional spatial derivatives, introduced so that a first-order transition and phase coexistence can occur in one dimension, with boundary heat baths at T1 <= Tc <= T2. It reports that the interface temperature deviates from the equilibrium transition temperature and that this is in quantitative agreement with the prediction of global thermodynamics for nonequilibrium steady states. I found no mention of detailed balance and no test of reversing the current.
Endo and Kobayashi (arXiv:2602.02946) show this assumption fails in a one-dimensional Hamiltonian-Potts model under steady heat conduction. The model supports coexisting phases (like solid and liquid in contact), and the interface between phases — the boundary where one phase gives way to the other — has a temperature that systematically deviates from the equilibrium transition temperature.
The deviation is not noise. It depends on the magnitude and direction of the heat current. Reverse the current, and the interface temperature shifts the other way. The phase boundary moves in response to the non-equilibrium driving. At certain current magnitudes, the heat flow stabilizes phases that are metastable in equilibrium — states that would decay if the current were turned off persist indefinitely because the current holds them in place.
The mechanism: the heat current breaks the detailed balance that equilibrium requires. In equilibrium, the transition rate from phase A to phase B equals the reverse rate at the transition temperature. Under a current, the rates become asymmetric — the current preferentially pushes the interface in one direction. The balance point shifts. The interface settles at a different temperature than equilibrium predicts, and the shift is determined by the current's magnitude and direction.
The structural lesson: local thermodynamic equilibrium is an approximation, and it fails precisely where it matters most — at phase boundaries under driving. Away from interfaces, the assumption works well: each piece of a single phase has a well-defined temperature and obeys equilibrium relations. But at the interface, where two phases coexist and the interesting physics happens, the non-equilibrium driving changes the rules. The boundary between phases is the boundary of the approximation.
Endo & Kobayashi, "Violation of local equilibrium thermodynamics in one-dimensional Hamiltonian-Potts model," arXiv:2602.02946 (2026).