Current flows downhill. A temperature gradient drives heat. A voltage gradient drives charge. A concentration gradient drives diffusion. The second law permits no other arrangement — or so it seems when you look at one channel at a time.
In coupled quantum dot systems, a current can flow against its own thermodynamic driving force. Particle current running from low concentration to high, or heat flowing from cold to hot, without any external work. This is the inverse current in coupled transport, and it does not violate the second law. The trick is that the system has multiple transport channels, and the entropy produced in the other channels more than compensates for the entropy consumed by the uphill flow. The total entropy production remains positive. The bookkeeping is satisfied. But the local current, viewed in isolation, defies its own gradient.
Note (2026-11-09): I checked the abstract of this paper by Ghosh and Ghosh, which is a review of thermodynamics of quantum coupled transport in quantum dots (single-dot thermoelectric cross-effects, then a three-terminal coupled quantum dot geometry). It discusses inverse currents in coupled transport, where a current flows against mutually parallel thermodynamic forces without violating the second law, and says they require breaking the symmetry between energy and particle transport, with conditions given for coupled dots with attractive interdot interactions. The essay presents this as a finding and adds details (Onsager symmetry in a single dot, entropy compensation in other channels) that the abstract does not state.
The mechanism requires breaking the symmetry between energy and particle transport. In a single quantum dot, the Onsager reciprocal relations tie particle and heat currents together with specific symmetry constraints. In coupled quantum dots with attractive interdot interactions, this symmetry breaks. The interaction between dots creates a new transport pathway that doesn't exist in either dot alone. The uphill current is not forced against the gradient by external work — it is permitted by coupling to a channel that shouldn't logically exist from the perspective of the individual components.
Each dot, analyzed in isolation, obeys its local gradient. The violation becomes possible only when the dots interact. The thermodynamic force still acts on each dot individually. But the coupled system accesses states that no analysis of individual components would predict. The whole doesn't just exceed the sum of the parts. It moves in a direction the parts cannot.