Colder Than Its Helpers: Quantum Coherence as a Cooling Resource

Can a quantum system become colder than the particles used to cool it? A new University of Toronto preprint shows how the answer can be yes—provided those particles carry something an ordinary thermal bath does not: carefully prepared quantum coherence. The result is a proposal for controlling small quantum systems, not a refrigerator that runs without resources. [1]

Submitted on October 7, 2026, Coherence-assisted quantum state preparation and thermal control through repeated interactions, by Alessandro Prositto and Dvira Segal, analyzes a qubit meeting a succession of prepared helper qubits. Its central advance is an exact description of how environmental coherence changes both the final state and the direction of energy exchange. The analysis does not assume that this coherence is only a tiny perturbation.

A bath made of individual quantum encounters

Ordinarily, a heat bath is imagined as a large surrounding environment: air around a device, water around a pipe, or vibrations inside a solid. A collision model replaces that complicated environment with a stream of small, controllable pieces. One helper interacts with the system, leaves, and is replaced by a freshly prepared helper. These auxiliary systems are called ancillas. Collision models are an established framework for studying open quantum dynamics and thermodynamics. [2]

In this study, both the system and each ancilla have two energy levels. Their energy gaps match. They exchange an excitation through a “flip-flop” interaction: an excited system can give its excitation to a helper, or receive one in the reverse process. Crucially, the ideal interaction conserves their combined bare energy. [1, Section II]

Incoming ancillas are identical and initially uncorrelated. No helper returns for another encounter, and helpers do not interact with one another. This is the memoryless, or Markovian, version of the model. The same encounter is repeated, but the analysis does not rely on a many-body material, an elaborate heat-engine cycle, or a hidden network of reservoirs.

The missing ingredient is phase, not just temperature

A thermal qubit is described by probabilities of occupying its lower and upper energy levels. For a positive temperature, the lower level is more populated. If the energy gap is fixed, the balance of those populations determines a temperature.

A coherent qubit contains additional information: a phase relationship between its energy states. Two helpers can therefore have the same populations, the same average energy, and the same population-defined temperature while differing in coherence. That difference matters when they interact with another quantum system.

With thermal, incoherent ancillas, the relaxing system approaches the ancillas’ populations and loses its own coherence. Prositto and Segal show that coherent ancillas change this simple picture. Population and coherence dynamics become coupled, opening access to stationary states that the corresponding thermal stream cannot prepare. Their exact solution accommodates physically allowed coherence strengths rather than treating all coherence effects as small corrections. [1, Section II]

Our reading of the result: an engineered environment is specified not only by how much energy it carries, but also by the quantum organization of that energy.

What “colder than the bath” actually means

The paper uses a precise definition of cooling: the system’s eventual probability of occupying its ground state exceeds the corresponding probability in each incoming ancilla. With equal energy gaps, that corresponds to a lower temperature inferred from populations. [1, Section III]

There is an important qualification. A coherent state is not an ordinary Gibbs thermal state. Calling it colder describes its populations, not every aspect of its quantum state. The authors therefore separate population preparation from a later step that can remove residual coherence.

A calculated example makes the distinction concrete

Using the paper’s steady-state expression, take an incoming ground-state probability of 0.80 and a coherence magnitude of 0.30. Set the dimensionless interaction parameters to 2Jτ = 0.20 and ωτ = 1.00, where J measures coupling, τ is encounter duration, and ω is the qubit frequency in the paper’s units. The predicted stationary ground-state probability is approximately 0.8624.

Converting the population ratios into temperatures gives a system temperature approximately 0.755 times the ancilla’s population-defined temperature—about 24.5% lower. We calculated these values from the preprint’s equations 11–15 and 31–32. They are not a universal performance figure, an energy-efficiency percentage, or a prediction for a particular laboratory refrigerator. [1]

The parameters are essential. Environmental coherence does not automatically cool the system. Depending on the coupling, encounter duration, and qubit frequency, it can instead increase heating. The analytical solution is useful precisely because it identifies where coherence helps a chosen task and where the same resource works against it.

Purity and cooling are different targets

Another result concerns purity, a measure of how mixed or well-defined a quantum state is. Lower energy does not necessarily mean a purer state, and higher purity does not necessarily mean lower energy. Coherence can change purity without changing populations in the same way.

In the cooling regime studied here, the final system is purer than it would be with incoherent ancillas having the same populations. But the converse fails: improved purity can also appear without cooling, including in certain heating regimes. A device designer must therefore specify the target instead of treating “colder,” “purer,” and “more coherent” as interchangeable. [1, Section III.2]

The authors also prove a limit: the stationary system purity cannot exceed the incoming ancilla purity within this model. The environmental resource improves what is reachable, but does not become an unlimited purification supply. This is a bound on the steady state, not a claim that an arbitrarily chosen initial system could never start purer than its helpers.

A phase dial for energy exchange

The thermodynamic analysis separates energy transferred during each encounter into two contributions. One comes from the population imbalance, reproducing the familiar tendency toward thermal equilibration. The other depends on coherence in both the system and the ancilla. [1, Section IV]

For fixed interaction parameters, the sign of this coherence-dependent contribution is controlled by the relative phase between those coherences. Their magnitudes set its size. A useful analogy is pushing a swing: changing timing can reverse an effect even when the strength of the push stays the same. Here, however, the relevant timing is encoded in quantum phase.

If the coherent contribution outweighs the conventional one, the total transfer can run against the direction suggested by population temperatures. Transfer can also occur when the two population distributions initially match, provided the coherence-dependent term is nonzero. Equal population temperatures are therefore insufficient to predict energy exchange between these nonequilibrium quantum states.

Another boundary matters just as much: at the stationary state, the model’s net heat and work currents vanish. Its lower population temperature is not evidence of an indefinitely operating refrigerator removing heat from an attached load. Connecting a load would change the problem and require a new analysis.

Why this is not a loophole in Carnot’s limit

The ideal collisions have zero net switching work under the paper’s interaction-energy accounting. That is a sharply defined statement about this exchange step. It does not mean that a laboratory can manufacture, synchronize, replace, and reset coherent ancillas for free. [1, Section IV]

Coherence makes the incoming stream a nonequilibrium resource. Looking only at its population temperature discards the very ingredient responsible for the unusual behavior. The second law must be applied to the full resource accounting, not to a stripped-down description in which the prepared coherence disappears from the ledger.

Carnot efficiency concerns a cyclic heat engine operating between thermal reservoirs under specified conditions. This paper does not report such an engine’s efficiency. It investigates state preparation and energy transfer using a nonthermal stream. “Beyond ordinary thermalization” is an accurate description; “proof that a heat engine beats Carnot for free” is not.

For practical quantum-energy research, that distinction is constructive rather than dismissive. The engineering question is whether the cost of supplying coherence is justified by better state preparation, control, or useful operation elsewhere in a device. That comparison remains to be demonstrated for an implementation.

From coherent preparation to an actual thermal state

Some applications need a thermal state, not merely populations associated with a desired temperature. The paper proposes a second stage: pure dephasing. This removes the phase relationships while preserving the ground- and excited-state probabilities. Afterward, the system is diagonal in its energy basis and can represent the intended thermal state. [1, Section III.3]

The protocol therefore divides the job into two parts: use coherent helpers to set the populations, then erase the leftover coherence without undoing that population setting. Such a sequence could be relevant to preparing inputs for quantum simulations, where a reproducible finite-temperature starting state matters.

Speed remains an open practical consideration. In the examples analyzed in the preprint’s convergence study, increasing ancillary coherence generally increases the number of encounters needed to approach the target. A more desirable final state need not arrive faster. Preparation time, accumulated hardware errors, and reset throughput all belong alongside final-state quality in an experimental comparison. [1, Appendix E]

The connection to Floquet engineering

This is primarily a repeated-interaction thermodynamics paper, not a demonstration of a new Floquet material. Its relevance to Floquet research lies in the control problem: how a recurring sequence of interactions changes the states a quantum system can reach.

Floquet engineering often emphasizes the timing, strength, and phase of external driving. This study offers a complementary lesson: the quantum state of the environment can be another control variable. Repeating a carefully chosen interaction is not equivalent to repeating that interaction with differently prepared environmental partners.

Our interpretation is that a useful future comparison would combine periodic system control with explicitly modeled ancilla preparation. Researchers could then separate the benefits of the drive from the benefits of environmental coherence, while accounting for both. The present paper supplies an analytically tractable baseline; it does not establish that such a combined device would improve energy efficiency.

What a convincing experiment should establish

A hardware test would need more than a lower excitation probability in a single run. The following checks would turn the proposal into a discriminating test of its mechanism:

• Compare coherent and incoherent helper streams with matched populations, energy gaps, coupling, and encounter duration.
• Measure both system populations and coherence, distinguishing population cooling from purification and from full thermal-state preparation.
• Sweep the relative phase to isolate the predicted coherence-dependent contribution to energy exchange.
• Track convergence and errors over repeated encounters, rather than reporting only the most favorable transient.
• Include helper preparation, resetting, synchronization, and any load when evaluating whole-device costs or cooling performance.

The bottom line: the Toronto preprint makes environmental coherence a precisely analyzable control resource. It shows how an ideal qubit can reach populations colder than its helpers, why phase can redirect energy transfer, and which purity and steady-state limits remain. Those are useful steps toward programmable quantum thermodynamics—not a finished energy technology and not an exemption from its accounting rules.