Heat flows from hot to cold. This is not a law you prove from first principles and then apply; it is the observation from which thermodynamics begins. The second law guarantees it in equilibrium. In steady-state transport between reservoirs, the Landauer-Büttiker formalism confirms it: heat conductance is positive, current flows down the temperature gradient.
Menczel, Flindt, and colleagues (arXiv:2602.21190) find that this can reverse during transients. In a quantum system coupled to reservoirs at different temperatures, the heat conductance can be transiently negative — heat flowing from cold to hot — depending on the initial preparation of the system. The effect requires strong coupling and non-Markovian dynamics, both captured exactly by their method.
The reversal is not a violation of thermodynamics. It is a transient powered by the initial state's correlations with the reservoirs. The system was prepared in a state that stores energy in system-reservoir correlations, and the relaxation of those correlations can drive heat temporarily against the gradient. Once the initial correlations decay, the conductance becomes positive and stays positive.
The distinction matters: steady-state heat conductance is always positive; transient heat conductance can be negative. The sign of heat flow is not a property of the temperature difference alone — it is a property of the temperature difference AND the initial conditions. In the Markovian, weak-coupling limit, the initial conditions don't matter and heat always flows correctly. Strong coupling and memory make history relevant.
The general observation: the direction of a macroscopic flow is fixed by gradients only in steady state. During transients, the initial preparation of the system can override the gradient. History competes with gradient, and for a time, history wins.
Author note (2026-11-06). The paper cited here, arXiv:2602.21190 (“Exact quantum transport in non-Markovian open Gaussian systems”), is by Guglielmo Pellitteri, Vittorio Giovannetti, Vasco Cavina. The name(s) this essay attributes it to (Menczel and Flindt) do not appear among the listed authors, so the attribution is wrong or refers to someone else's earlier work; I have not worked out which. Found by a mechanical check of the names that precede 'and colleagues' (or 'and collaborators') in the essay's opening paragraphs against the arXiv author list; the rest of the essay was not re-checked in this pass.