A Quantum Maxwell’s Demon Scales Information into Energy Flow

Maxwell’s demon is usually introduced as a paradox: an imaginary being watches molecules, opens a tiny door at just the right time, and appears to make heat flow the “wrong” way. Modern quantum thermodynamics treats the demon less as a fantasy and more as a design principle. Information, if acquired and used correctly, can become a thermodynamic resource.

A new preprint by Shotaro Oki, Yuki Kadono, Kaito Tojo, Takahiro Sagawa and Ken Funo, “Quantum enhancement of information-mediated energy transfer”, posted to arXiv on June 26, 2026, asks a timely energy question: can quantum coherence and collective coupling make information-powered energy transfer scale better, without paying the usual dissipation cost? Their answer is a carefully qualified yes. In a theoretical open quantum system, collective jumps can make heat currents and information flow grow linearly with the number of degenerate states, while entropy production remains order-one.

The headline is not “free energy from information.” It is subtler and more useful: quantum collectivity can raise the power ceiling of an information engine without making entropy production scale at the same rate.

Why this belongs in beyond-Carnot thermodynamics

Carnot efficiency is the classic limit for heat engines operating between two thermal reservoirs. But many quantum-energy proposals do not look like nineteenth-century piston engines. They involve measurement, feedback, coherence, correlations, engineered reservoirs, periodic driving and strongly structured spectra. These ingredients do not abolish the second law; instead, they force researchers to write the law with more terms.

One of those terms is information flow. When two subsystems share quantum mutual information, the thermodynamic accounting for either subsystem alone can look strange. A controlled subsystem may appear to pump heat against a temperature gradient, or extract more work than its local free-energy change would suggest, because another subsystem is effectively acting as a demon. The full composite system still obeys the second law, but the local balance sheet includes information as a resource.

Oki and colleagues place this logic inside the language of open quantum systems. They consider two interacting subsystems, X and Y, each weakly coupled to its own heat baths and evolving under a Gorini–Kossakowski–Sudarshan–Lindblad master equation. The framework is not a laboratory experiment yet; it is a theory paper that builds a scaling argument and a concrete model. But the subject is practical: how to move energy in small quantum devices with high power and low waste.

The new bound: a thermodynamic uncertainty relation for information

The paper’s first step is a short-time thermodynamic uncertainty relation, or TUR. In ordinary nonequilibrium physics, TURs express a trade-off: precise, large currents generally require entropy production. They are valuable because they turn the second law from a qualitative warning into a quantitative design constraint.

Here the current of interest is not only heat current but information flow. At nonequilibrium steady state, the authors derive a bound connecting the square of the information flow to the product of two quantities: partial entropy production and partial activity. The activity is roughly the average jump rate weighted by the energy of those jumps. In plainer language, it measures how busy the quantum transition network is.

This is where the quantum advantage enters. If many transitions merely add independently, activity scales like N. But if the system is collectively coupled to a bath, jump amplitudes can add coherently. That is the same broad physical family as Dicke superradiance, where many emitters radiate collectively rather than as isolated atoms. In the model, collective jump amplitudes scale as N, so the activity can scale as N². The TUR then permits information flow that scales as N even when entropy production does not.

An autonomous quantum Maxwell’s demon

The concrete engine in the paper uses two subsystems with N-fold degenerate ground and excited manifolds. The subsystems interact, but the interaction commutes with their local Hamiltonians. This “bipartite” condition lets the model behave like an autonomous Maxwell’s demon: no external agent is clicking a measurement button; the information processing is built into the coupled dynamics.

Subsystem X is coupled to one cold bath and acts as the demon. Subsystem Y is coupled to two baths with different temperatures and acts as the feedback-controlled system. The authors choose energy-selective couplings so that the four relevant collective states form a closed thermodynamic cycle. Under a suitable positive cycle affinity, the collective transition rates drive the cycle in the direction associated with demon operation.

The result is the central scaling claim. Information flow from Y to X scales as O(N). That information flow sustains an O(N) reverse heat current from the colder bath to the hotter bath. Yet the partial entropy production rates for X and Y remain O(1). In a non-collective version, making the same O(N) information flow would require entropy production to scale up as well. Collective quantum mechanics changes the engineering trade-off.

The demon does not violate thermodynamics; it changes which resource is doing the work. The resource is collectively enhanced information flow, paid for within the total entropy balance.

Why the scaling matters for quantum energy devices

For a smart non-physicist, the word “scaling” may sound abstract. In energy technology it is everything. A mechanism that works beautifully for one atom but collapses when expanded to ten or one hundred components is rarely useful. A mechanism whose useful output grows with system size while its waste grows more slowly deserves attention.

The proposed demon is not ready to wire into a device. It assumes structured degeneracies, carefully engineered collective couplings, weak system-bath interactions and the approximations normally used to derive a Markovian master equation. Those are strong assumptions. But they are not alien to current quantum platforms. Superconducting circuits, trapped ions, cavity QED, Rydberg arrays and solid-state spin ensembles all already use collective modes, engineered reservoirs and selective transitions.

The relevance to Floquet engineering is indirect but important. Floquet systems are controlled by periodic Hamiltonians; quantum thermodynamic machines are controlled by energy exchanges. In many proposed heat engines, batteries and thermal routers, periodic driving creates the sidebands or dressed states through which energy flows. If information-flow TURs become a usable design language, they could help engineers decide when a periodic protocol is merely moving heat around and when it is creating scalable, low-dissipation information-mediated transport.

Beyond steady state: a quantum-battery connection

The authors also look beyond nonequilibrium steady states. In steady operation, the information flow in their demon is dissipative: it is generated by bath-induced jumps. Away from steady state, there is also a unitary contribution generated by the interaction Hamiltonian. Collective interactions can amplify this unitary component.

Applied to nonequilibrium free-energy charging of a quantum battery, the paper reports an O(N²) enhancement of the free-energy charging power. That is a different mechanism from the reverse-heat-current demon, but it points in the same direction: collective quantum structure can turn information flow into a power resource. Quantum batteries are often judged by stored energy, charging power and ergotropy, the portion of stored energy extractable as useful work. The new paper adds another lens: how quickly free energy can be delivered when correlations and mutual information are engineered rather than ignored.

What would make this experimental?

Turning the proposal into an experiment would require three ingredients. First, a platform with near-degenerate manifolds that can be addressed collectively. Second, heat-bath couplings that are energy selective enough to implement the intended transition cycle. Third, measurement methods that can infer heat currents, entropy production and information flow without destroying the correlations that make the device work.

None of these is trivial. Measuring heat at the quantum scale remains difficult, as recent superconducting-qubit calorimetry work illustrates. Verifying information flow is also subtler than measuring an average energy current, because it depends on correlations between subsystems. The cleanest early demonstrations may therefore be quantum simulators, where the “baths” are engineered and the relevant states can be reconstructed with high control.

There is also a materials question. Collective enhancement is powerful only if coherence survives. Disorder, dephasing and uncontrolled bath channels could turn the desired N² activity back into ordinary additive noise. This is a familiar theme across Floquet materials and quantum batteries: the resource is not energy alone, but phase-coherent, structured energy.

The bigger lesson

Beyond-Carnot research is sometimes misunderstood as an attempt to beat the second law. The better interpretation is that quantum systems force us to identify every resource in the balance sheet: heat, work, coherence, information, correlations, activity and time-dependent control. Once those resources are explicit, new performance regimes become possible without magic.

Oki, Kadono, Tojo, Sagawa and Funo have added an important piece to that balance sheet. Their theory says that information-mediated energy transfer has its own uncertainty relation, and that collective quantum jumps can relax the usual power-dissipation trade-off. For Floquet and quantum-energy researchers, that is a useful design message: the best future machines may not simply drive harder. They may coordinate more intelligently.

Sources and further reading

• Shotaro Oki, Yuki Kadono, Kaito Tojo, Takahiro Sagawa and Ken Funo, “Quantum enhancement of information-mediated energy transfer”, arXiv:2606.27868 (2026).
• Takahiro Sagawa and Masahito Ueda, “Second Law of Thermodynamics with Discrete Quantum Feedback Control”, Physical Review Letters 100, 080403 (2008).
• Jordan M. Horowitz and Massimiliano Esposito, “Thermodynamics with Continuous Information Flow”, Physical Review X 4, 031015 (2014).
• Andre C. Barato and Udo Seifert, “Thermodynamic Uncertainty Relation for Biomolecular Processes”, Physical Review Letters 114, 158101 (2015).
• Robert H. Dicke, “Coherence in Spontaneous Radiation Processes”, Physical Review 93, 99 (1954).
• Robert Alicki and Mark Fannes, “Entanglement boost for extractable work from ensembles of quantum batteries”, Physical Review E 87, 042123 (2013).