Idle-Level Quantum Heat Engine Architecture

• Idle-level quantum heat engine architecture is a design paradigm that partitions a quantum Otto engine's spectrum into working levels that perform work and idle levels that reshape heat currents.
• It leverages spectrally invariant idle states to redistribute thermal populations, thereby enhancing efficiency beyond the standard two-level qubit benchmarks without producing extra work.
• Implementations in qutrit and two-qubit non-Hermitian XY models demonstrate how independent control of idle-sector energetics can optimize engine performance while influencing work fluctuations.

Idle-level quantum heat engine architecture denotes a class of multilevel quantum Otto engines in which only a subset of the spectrum is driven by the external work parameter, while one or more thermally accessible eigenlevels remain spectrally unchanged during the adiabatic strokes. These unchanged states are the “idle levels.” Their defining role is not to generate work directly, but to reshape thermal populations and heat currents, thereby modifying efficiency relative to ordinary qubit-like Otto engines with uniformly scaled gaps. The concept was formulated explicitly for multilevel and interacting-spin Otto engines by de Oliveira and Jonathan, and later received a concrete microscopic realization in a two-qubit non-Hermitian XY model whose spectrum separates into working and idle sectors (Anka et al., 2021, Tusun et al., 21 Jun 2026).

1. Defining principle and thermodynamic structure

The canonical starting point is a quantum Otto cycle for a Hamiltonian

$$\mathcal{H}(\lambda)=\sum_n E_n(\lambda)\, |E_n(\lambda)\rangle \langle E_n(\lambda)|,$$

with an externally varied control parameter $\lambda$. In the standard two-bath version, the working substance thermalizes at $\lambda_i$ with a cold bath, undergoes an adiabatic stroke to $\lambda_f$, thermalizes at $\lambda_f$ with a hot bath, and is then adiabatically driven back to $\lambda_i$. Heat on the isochores is

$$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$

and total work is

$W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$

with $\Delta E_n = E_n^h-E_n^c$ and $\Delta p_n=p_n^h-p_n^c$. The efficiency is

$\lambda$0

An idle level is an eigenenergy that does not change when the work parameter is varied. In the simplest magnetic implementations this means

$\lambda$1

for part of the spectrum, while other levels retain explicit $\lambda$2-dependence. Such levels are thermally populated during the nonadiabatic strokes, so they contribute to $\lambda$3 and $\lambda$4, but because they do not shift during the adiabatic strokes they contribute nothing directly to work. Architecturally, the spectrum is partitioned into a work sector and an idle sector.

This immediately distinguishes idle-level engines from ordinary qubit Otto engines. If all gaps scale by one common ratio $\lambda$5, then the efficiency is rigidly fixed at

$\lambda$6

and for field-driven qubit examples this becomes

$\lambda$7

Idle levels break this proportional scaling, allowing heat-current engineering without altering the basic Otto sequencing (Anka et al., 2021).

2. Original multilevel Otto proposals

The original explicit idle-level architecture was developed in two working substances: a single qutrit and a pair of interacting spin-$\lambda$8 particles with XXZ Heisenberg coupling (Anka et al., 2021).

For the qutrit, the Hamiltonian is

$\lambda$9

with eigenvalues $\lambda_i$0, $\lambda_i$1, and $\lambda_i$2. The $\lambda_i$3 pair are the working levels because they shift with the field $\lambda_i$4, whereas $\lambda_i$5 is the idle level because it is independent of $\lambda_i$6. The resulting work depends only on the working levels,

$\lambda_i$7

so the $\lambda_i$8 level carries heat but not work.

For the coupled-spin realization, the Hamiltonian is

$\lambda_i$9

Its field-dependent levels are $\lambda_f$0 and $\lambda_f$1, while the interaction-generated middle levels

$\lambda_f$2

are $\lambda_f$3-independent and therefore idle. This realizes the same architecture in a composite spin system: a working manifold whose energies are dragged by the external field, and an idle manifold thermally populated but mechanically inactive under the work stroke.

The original paper treated both a standard two-bath Otto engine and a measurement-based Otto engine. In the latter, the hot isochore is replaced by a nonselective quantum measurement or, more generally, a unital quantum channel,

$\lambda_f$4

with injected energy

$\lambda_f$5

A key theorem in that framework is that if $\lambda_f$6 is unital and the pre-measurement state is passive, then the measurement injects nonnegative energy and can replace the hot bath consistently.

3. Mechanism of efficiency enhancement

Idle-level enhancement is a heat-current effect, not a direct work-gap enhancement. In the qutrit architecture, the paper distinguishes the $\lambda_f$7 pair as working levels and $\lambda_f$8 as the idle level. Because only the $\lambda_f$9 energies move during the adiabatic strokes, any energy exchanged through $\lambda_f$0 cannot be converted into work. Its role is instead to alter the denominator of the efficiency by redistributing heat among the isochores (Anka et al., 2021).

The original analysis shows that efficiency enhancement beyond the qubit benchmark $\lambda_f$1 requires the idle channel to mediate a reversed heat flux from the cold side to the hot side. In the qutrit case, this condition is expressed by the idle-level population change $\lambda_f$2. The same logic extends to the XXZ model: the idle manifold improves efficiency only when its net heat contribution is negative in the appropriate sign convention, meaning that the idle sector carries heat oppositely to the usual hot-to-cold direction. The gain does not arise from extra work production by the idle levels; it arises because part of the total absorbed energy is rerouted through a channel that does not participate in work extraction.

This point is especially important because the qutrit realization is indivisible. The effect therefore cannot be attributed to entanglement or multipartite correlations. The architecture is spectral and thermodynamic before it is many-body.

The measurement-based variant extends this logic by giving direct control over post-measurement populations. In the qutrit case, rank-1 projective measurements parameterized in an $\lambda_f$3 basis can redistribute population on the idle level more favorably than a thermal hot bath. This allows the measurement-driven engine to outperform the corresponding two-bath idle-level engine for suitable measurement bases. At the same time, the high-efficiency limit is singular: for one reported choice of measurement basis, the efficiency approaches $\lambda_f$4 as $\lambda_f$5, but the transferred heats and work simultaneously go to zero. The architecture can therefore approach reversible-looking behavior only in a trivial vanishing-output limit.

4. Fluctuations and reliability of idle-level engines

The next major development was the fluctuation analysis of a two-spin Heisenberg Otto engine with idle levels, where the Hamiltonian is

$\lambda_f$6

Its spectrum contains field-sensitive working levels at $\lambda_f$7, a field-insensitive level at $\lambda_f$8, and an idle level at $\lambda_f$9 (Anka et al., 2023).

The average work over one cycle is

$\lambda_i$0

with

$\lambda_i$1

and

$\lambda_i$2

A central result is that work and work fluctuations can be rewritten in terms of equilibrium magnetization and magnetic susceptibility: $\lambda_i$3 and

$\lambda_i$4

This connects idle-level thermodynamics directly to equilibrium response functions.

The fluctuation study showed that idle levels do not generically stabilize microscopic engines. Weak coupling can yield regions where the efficiency exceeds the uncoupled-spin Otto value

$\lambda_i$5

while still maintaining relative work fluctuations below unity, but only in a narrow regime with high $\lambda_i$6 and low $\lambda_i$7. In strong coupling, where the idle level can become the ground state, relative fluctuations remain large throughout the explored parameter range. The reported asymptotic minimum is

$\lambda_i$8

for $\lambda_i$9 and $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$0. The conclusion is that idle levels can enhance mean efficiency but do not make this microscopic engine a reliable work source.

5. Microscopic realization in a non-Hermitian XY model

A concrete microscopic realization of the idle-level architecture was provided in a two-qubit non-Hermitian XY model with a staggered imaginary magnetic field (Tusun et al., 21 Jun 2026). The Hamiltonian is

$$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$1

with $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$2 and $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$3. In the computational basis it is block diagonal, splitting into a working block and an idle block: $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$4

The corresponding eigenvalues are

$$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$5

for the working sector, and

$$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$6

for the idle sector. This realizes the architecture exactly: $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$7 move with the work parameter $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$8, whereas $$Q_c = \sum_n E_n(\lambda_i)(p_n^c - p_n^h), \qquad Q_h = \sum_n E_n(\lambda_f)(p_n^h - p_n^c),$$9 are independent of $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$0 and therefore idle during the adiabatic strokes.

The Otto work formula is

$W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$1

where $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$2, $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$3, and $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$4. The key structural point is that the work numerator depends on the moving working levels, while the non-Hermitian control $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$5 enters through partition-function denominators containing $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$6. Idle-level tuning therefore modifies occupations and absorbed heat without directly changing the work-stroke gap.

For the reference parameters

$W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$7

the Carnot efficiency is $W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$8. The model exhibits a transition from a dissipative regime to genuine heat-engine operation at

$W = -(Q_h+Q_c)= -\sum_n \Delta E_n \Delta p_n,$9

For $\Delta E_n = E_n^h-E_n^c$0, the engine has $\Delta E_n = E_n^h-E_n^c$1 and $\Delta E_n = E_n^h-E_n^c$2. The efficiency rises from about $\Delta E_n = E_n^h-E_n^c$3 at $\Delta E_n = E_n^h-E_n^c$4 to $\Delta E_n = E_n^h-E_n^c$5 at $\Delta E_n = E_n^h-E_n^c$6, and approaches

$\Delta E_n = E_n^h-E_n^c$7

as $\Delta E_n = E_n^h-E_n^c$8, still below Carnot. In the Hermitian limit $\Delta E_n = E_n^h-E_n^c$9, the same parameter set yields

$\Delta p_n=p_n^h-p_n^c$0

placing the system in the dissipative regime. The paper’s interpretation is that non-Hermiticity is indispensable because it provides an independent control knob acting only on the idle-level gap.

6. Limitations, misconceptions, and architectural significance

Idle-level quantum heat engines do not derive their advantage from a generic quantum enhancement of Otto efficiency, nor do they evade conventional thermodynamic bounds. In the original formulation, the relevant comparison is with the rigid qubit value $\Delta p_n=p_n^h-p_n^c$1, and the enhancement comes from heat-current repartition through levels that remain invariant under the work parameter. Carnot is approached only in vanishing-output limits in the measurement-based qutrit case, and the non-Hermitian XY realization remains below Carnot even when the idle gap is maximally compressed (Anka et al., 2021, Tusun et al., 21 Jun 2026).

Another common misconception is to identify idle-level performance with entanglement-assisted work extraction. The original qutrit model already rules this out, because it is indivisible and still exhibits the idle-level effect. In the coupled-spin models, interaction is important because it creates $\Delta p_n=p_n^h-p_n^c$2-independent dressed levels, not because entanglement by itself is the thermodynamic resource.

Architecturally, the central design rule is consistent across the literature: a useful idle-level engine must separate the spectrum into a work sector and an idle sector, keep the idle sector thermally accessible, and maintain independent control over the idle-sector energetics or populations. In the original proposals this is achieved by spectrum topology and population routing; in the non-Hermitian XY model it is achieved by an independently tunable idle-gap parameter $\Delta p_n=p_n^h-p_n^c$3; in the fluctuation study it appears as the coupling-dependent redistribution of Gibbs weight into a non-working level. The same studies also show the main limitation of the architecture: idle levels can improve mean efficiency, but they can just as easily trap thermal population, reduce mean work, and amplify relative fluctuations.

Taken together, the literature establishes idle-level quantum heat engine architecture as a distinct Otto-engine design principle. Its hallmark is not stronger work extraction per active level, but the deliberate inclusion of thermally active, mechanically inert spectral channels that reshape $\Delta p_n=p_n^h-p_n^c$4, $\Delta p_n=p_n^h-p_n^c$5, and $\Delta p_n=p_n^h-p_n^c$6. In that sense, idle levels are neither mere spectators nor auxiliary decoration. They are a controlled thermodynamic substructure whose value depends on how effectively the architecture converts spectral decoupling into favorable heat-current routing.