A quantum work-extraction proposal puts complementary measurements to work

Information as a thermodynamic resource

Chellasamy Jebarathinam, Debarshi Das and Ray-Kuang Lee investigate a cooperative arrangement involving two observers, Alice and Bob. They share a quantum state. Alice measures her part and sends the outcome to Bob through an ordinary classical communication channel. Bob uses that information to choose how to extract work from his own system.

The study, submitted to arXiv on November 10, 2026, is a theoretical preprint, not an experimental demonstration.

Bob's permitted actions are described by the resource theory of thermal operations. He can interact with an environment at a specified temperature, but the combined transformations must conserve energy. Within the paper's asymptotic description—appropriate to processing many copies—the available work is expressed through the system's departure from thermal equilibrium.

Alice's message adds another contribution. Knowing her outcome lets Bob distinguish between conditional states that would otherwise be averaged together. The paper expresses this information-related contribution using the Holevo quantity: an upper bound on the information accessible about Alice's outcomes from Bob's quantum states.

This is an accounting of assisted work on Bob's side. It does not establish the net output of a complete engine after accounting for the helper's operations and communication.

Why use more than one measurement basis?

The central proposal divides Bob's particles into separate pools. Each pool receives systems associated with a different measurement setting on Alice's side and undergoes an appropriate work-extraction operation.

The settings use mutually unbiased bases. These are complementary ways of measuring a quantum system: a state with a definite outcome in one basis gives equally likely outcomes when measured in another. The question is whether correlations remain useful across these different perspectives.

The authors distinguish two strategies. One chooses complementary settings together to maximize the weakest information contribution among them. Another starts with the basis providing the strongest classical correlation, then looks for information remaining in complementary bases.

In the latter strategy, the paper labels the first contribution C1 and the residual contributions Q2 and Q3. For two pools containing M and N particles, their relative size is c = N/M. The information-related part of the average work is then proportional to the weighted expression (C1 + cQ2)/(1 + c). Adding a third pool introduces another weight and another residual contribution.

Importantly, these measurements concern different ensembles. The protocol does not extract several complementary measurement outcomes from the same particle.

What would count as a quantum advantage?

The proposed tests compare assisted work against local hidden-state models. Such models explain Bob's conditional states through a pre-existing ensemble of local quantum states, with Alice's outputs correlated with a hidden variable. Excluding that explanation certifies quantum steering.

In the standard trust arrangement, Bob's system is characterized while Alice's measurement device is treated as a black box. The authors propose work-based inequalities intended to certify steering without trusting the internal operation of Alice's apparatus.

They also consider a stronger assumption: Alice's system is known to be a qubit, a two-dimensional quantum system. This changes the permitted classical explanation. The dimension-restricted framework also limits the hidden-variable resources available to simulate the correlations.

Under that restriction, the authors propose adaptive benchmarks whose information-related terms decrease as 1/(1 + c) for two pools and 1/(1 + c + d) for three. They argue that suitable correlations can exceed these benchmarks even when ordinary steering tests do not reveal an advantage. The associated resource is called superunsteerability: difficulty reproducing correlations with the restricted hidden-variable model, rather than a violation of every unrestricted classical explanation.

A proposal with unresolved numerical details

The distinction between these trust frameworks matters. A lower dimension-restricted benchmark is not evidence that changing assumptions creates energy. It changes which explanations an observed work value could exclude.

The source also contains inconsistent numerical illustrations. For its three-setting example, it gives both approximately 1.91 and 2.33 as the relevant pool-ratio crossover in different passages. Those values should not be treated as an established operating threshold.

The detailed analytical examples emphasize two-qubit Bell-diagonal states. The paper discusses possible superconducting and photonic implementations, but reports no implementation of its proposed protocol.

Its central idea is a connection between complementary correlations, local work extraction and explicit trust assumptions. Establishing a dependable experimental witness requires resolving the quantitative inconsistencies and validating the claimed bounds for the intended setting.