Quantum computers like the cold and produce heat, which is an awkward combination: superposition and entanglement wash out as soon as the environment warms up. A paper in Physical Review Letters proposes turning that around by using one of quantum mechanics' least tangible features to make the cooling itself more efficient.
The task, devised by an international team led by Tanmoy Biswas at the University of Cologne, is a theoretical cooling protocol. An experimenter sacrifices an "assemblage" — a set of labelled quantum states — to pull heat out of a bath. The states must be finite-dimensional, but their form is otherwise unrestricted; nuclear spins in nitrogen-vacancy centres in diamond are one candidate. Every state in the set is distinct and carries its own label. Depending on the label of the chosen state, the experimenter drives the system's energy to a specified level. Seeking thermal equilibrium, the state has to absorb heat from the bath. It is then discarded and the cycle restarts with a fresh state.
Not every assemblage can do this. The ones that cool effectively share a property: they can be used to demonstrate a form of nonlocality called quantum steering. Steering occurs when the outcome of one party's measurement on their half of an entangled state influences the state the other party holds afterwards. Erwin Schrödinger described it in 1935; demonstrating it requires both shared entanglement and incompatible measurements, which makes it a distinctly quantum resource rather than a statistical curiosity.
In the team's setup, Alice randomly picks a measurement to perform on her half of the entangled pair — rolling dice first, if she likes. Unbeknownst to Bob, his system then assumes a particular form. Alice acts as a virtual adviser, contacting Bob (perhaps by phone) to reveal her measurement choice and outcome. Together these specify Bob's post-measurement state and tell him what to do next. Counterintuitively, Alice has no access to the bath at all, yet she helps prepare the assemblage used to cool it.
The result is a strict separation: assemblages that demonstrate steering give an advantage in the cooling task over those exhibiting only classical correlations. "Our goal was not simply to demonstrate an advantage for a particular quantum state, but to establish a universal connection between quantum steering and thermodynamic performance," Biswas says.
Steering already has a resume — powering heat engines, certifying entanglement, generating randomness. The cooling task here is not cryptographic; neither party cares about secrecy, and entanglement participates only indirectly. What remains open is which specific operations and assemblages are optimal, a question the paper leaves for others.
The researchers hope the result finds use in the platforms people actually build quantum computers from, trapped ions and superconducting qubits. For Biswas, though, the headline is more fundamental: "every steerable quantum correlation, no matter how weak, can provide a genuine thermodynamic advantage in cooling" — a direct line from a 90-year-old piece of quantum foundations to a practical thermodynamic task.
