Pudim AI Story / STORY-2026-0006

When a Quantum Swap Is Hard to Improve

A finite bound shows when a simple swap between two thermal oscillators captures nearly all available work, leaving an exponentially small reward for more elaborate control.

Research Program
Quantum ThermodynamicsMathematical Physics
An ivory paper bell at upper left and a green bell at lower right frame mustard discs and ivory squares on a deep plum background.
Two paper bells suggest different natural frequencies, while colored inserts suggest an exchange. Conceptual illustration of a mode swap. Original conceptual illustration generated for Pudim Physics · CC-BY-4.0

How much is better control worth?

A temperature difference offers a way to extract work. Yet a small quantum system can hold energy that even an elaborate controller cannot recover. How can we tell when a simple operation already gets almost everything available?

A new Pudim Letter answers this question for two quantum oscillators prepared independently at different temperatures. Each oscillator stores excitations in equally spaced energy steps. Its natural frequency sets the size of those steps, so moving an excitation between oscillators can change the total energy.

The paper bells above suggest two different natural notes and an exchange. In the calculation, a mode swap exchanges the oscillators' excitation distributions while their frequencies stay fixed. The bells are a visual metaphor; no experiment with bells is proposed.

Earlier research had already identified when a swap extracts work and when it is best within a standard family of oscillator controls. The new result bounds the extra work available to every more general operation in the stated model. It places a ceiling on the reward for added control.

Scientific traceRO-2026-0008:C1RO-2026-0008:C2

What earlier work established

In 2016, Eric Brown, Nicolai Friis and Marcus Huber studied controls that mix oscillator modes and reshape their fluctuations, a family called Gaussian operations. They showed how work can remain accessible through highly excited states after these restricted controls are exhausted. Their analysis did not give the finite bound developed here.

In 2021, Massimiliano Sacchi calculated work from a swap protocol and described its operating window. That established when the simple exchange can be useful. It did not bound what every possible alternative operation could achieve.

In 2025, Frank Ernesto Quintela Rodriguez and colleagues developed general results for Gaussian work extraction and estimates based on entropy. This is the closest quantitative benchmark. The new Letter sharpens the allowance for additional work by examining the detailed ordering of energies and probabilities.

Scientific traceRO-2026-0008:C1RO-2026-0008:C5RO-2026-0008:C3RO-2026-0008:C6

The theorem and its boundary

The model considers a closed pair of oscillators during one extraction stroke. An allowed operation changes the states making up the quantum mixture while keeping their statistical weights unchanged. Those states can carry different energies afterward, so measured energy probabilities can change. With the coupling off at both endpoints, the decrease in the pair's average energy counts as extracted work.

For a swap to lower the energy, the hotter oscillator must have the higher frequency. Its frequency ratio relative to the colder oscillator must also remain below their temperature ratio. These two conditions define the known operating window; the frequency mismatch is called detuning.

The striking conclusion concerns a particular approach to equilibrium. The temperature imbalance tends to zero while detuning shrinks with it, staying at a fixed position inside that window. The background temperature scale and the lower oscillator frequency remain fixed.

Inside the window, swap work has a leading size proportional to the square of the temperature imbalance. The theorem bounds every possible improvement by an exponentially small quantity. That bound shrinks faster than any fixed power of the imbalance, so the swap captures an ever larger fraction of the available work.

The result also supplies a bound at finite parameter values. In one worked example, the temperature difference is one fortieth of their sum, and detuning lies one quarter of the way across the window. At the chosen temperature scale, any improvement is less than four millionths of the swap's work. This is an upper bound on added work, not a measured yield.

Scientific traceRO-2026-0008:C1RO-2026-0008:C2RO-2026-0008:C3

Why little work remains

The proof groups states by their total number of excitations. After the swap, a whole sequence of low-excitation groups is already in the most favorable order. Their energy ranges stay ordered, and the more probable states occupy the lower energies. Rearranging these groups cannot release more work.

Any further gain must therefore come from states beyond that ordered sequence. Such highly excited states occur rarely in the thermal preparations. The proof bounds the energy they contain, which also bounds all work that any further control could extract from them.

As the temperature imbalance and detuning shrink together, the ordered sequence extends to higher excitations. The remaining probability falls exponentially. This reasoning controls the infinite set of possible states and yields a directly calculable finite bound, rather than relying on a large numerical cutoff alone.

Scientific traceRO-2026-0008:C1RO-2026-0008:C2

How much energy is accessible?

An established thermodynamic estimate compares the initial energy with that of an equilibrium state having the same entropy. Entropy measures the spread of those mixture weights. The energy difference gives an allowance for extractable work, but preserving entropy alone imposes fewer restrictions than preserving all the weights.

In the stated limit, the fraction of that allowance recoverable by the best allowed control follows a parabola across the operating window. It rises from zero at one edge, reaches the full allowance at the midpoint to leading order, and falls to zero at the other edge. This fraction concerns the entropy-based allowance, not all the initial energy.

At equal frequencies, the swap extracts no work. More general control can still extract work, but its entire yield is exponentially small in the prescribed limit. The paper also determines that exponential rate at both window edges and outside them; a vanishing fraction does not mean exactly zero work.

A further result asks which excitation states a controller must be able to address. At equal frequencies, the minimum required range grows without bound as the temperature imbalance disappears. Inside the window, any chosen fraction below 100 percent of the leading work can be reached with a range that stays finite as the imbalance shrinks.

Scientific traceRO-2026-0008:C3RO-2026-0008:C4RO-2026-0008:C5

What the classical calculations check

The accompanying numerical study tests 84 parameter choices, including both edges of the window and regions outside it. It sorts energies and probabilities using high-precision arithmetic. These results are classical computation; the release reports no new physical experiment or evidence from a quantum processor.

A separate computational reproduction reran all 84 sorting cases and regenerated the three figures. Repeated calculations at greater precision check arithmetic stability. Analytic bounds control the energy omitted from the finite calculations. Numerical agreement does not replace the proof or rigorous control of rounding errors.

Scientific traceRO-2026-0008:C7

What the result does not settle

The result concerns two independently prepared thermal oscillators and a fixed-frequency extraction stroke. Preparing those states, tuning frequencies between preparations, and operating the controller can carry additional costs. The paper does not supply the efficiency of a complete engine or a guarantee about extraction speed and power.

The way equilibrium is approached also matters. Keeping a nonzero frequency mismatch fixed while reducing the temperature difference is a different limit. Likewise, the number of excitation states addressed by a controller is not a measure of its running time or practical complexity.

The bells return us to the simple exchange. Within the stated model and detuning window, the bound can show that almost no work remains for a more elaborate controller to recover. That makes the value of added control a quantitative question, with a useful answer before a new protocol is designed.

Scientific traceRO-2026-0008:C1RO-2026-0008:C2RO-2026-0008:C5
Open the paragraph-level evidence register

Every material assertion is bound to accepted claims and evidence in the immutable source release.

p1 · interpretationRO-2026-0008:C1RO-2026-0008:C2
p2 · interpretationRO-2026-0008:C1
p3 · analogyRO-2026-0008:C1
p4 · resultRO-2026-0008:C1
p5 · interpretationRO-2026-0008:C1RO-2026-0008:C5
p6 · interpretationRO-2026-0008:C1RO-2026-0008:C3
p7 · interpretationRO-2026-0008:C1RO-2026-0008:C6
p8 · limitationRO-2026-0008:C1
p9 · interpretationRO-2026-0008:C1RO-2026-0008:C2
p10 · limitationRO-2026-0008:C2
p11 · resultRO-2026-0008:C2RO-2026-0008:C3
p12 · resultRO-2026-0008:C1RO-2026-0008:C2
p13 · interpretationRO-2026-0008:C1
p14 · interpretationRO-2026-0008:C1
p15 · interpretationRO-2026-0008:C1RO-2026-0008:C2
p16 · interpretationRO-2026-0008:C3
p17 · resultRO-2026-0008:C3
p18 · resultRO-2026-0008:C4
p19 · resultRO-2026-0008:C5
p20 · resultRO-2026-0008:C7
p21 · resultRO-2026-0008:C7
p22 · limitationRO-2026-0008:C1RO-2026-0008:C2RO-2026-0008:C5
p23 · limitationRO-2026-0008:C2RO-2026-0008:C5
p24 · interpretationRO-2026-0008:C1RO-2026-0008:C2