Particle-Exchange Quantum Heat Engine

• Particle-Exchange Quantum Heat Engine is a class of quantum thermal machines that use selective transfer of discrete carriers (electrons, photons, or atomic spins) to convert heat into work.
• These engines employ energy filtering via quantum dots, nanowires, or cavity modes to achieve high efficiency and steady-state operation in miniaturized setups.
• Performance is influenced by quantum coherence, interference, and many-body effects, enabling practical implementations across electronic, atomic, and photonic architectures.

Particle-exchange quantum heat engine denotes a class of quantum thermal machines in which useful work is produced by selective exchange of discrete carriers between reservoirs rather than by a classical cyclic working fluid with moving parts. In one prominent realization, a temperature bias drives electrons through an energy-selective nanoscale conductor against an electrical bias; in others, the exchanged quanta are photons emitted into a cavity mode or spin quanta transferred in inelastic atomic collisions. The common thermodynamic structure is that heat is absorbed from a high-energy source, part of that energy is converted into work, and the remainder is rejected or dissipated through imperfect filtering, finite linewidth, interaction effects, or irreversible machine–bath contact (Josefsson et al., 2017, Bouton et al., 2020, Sarmah et al., 2023).

1. Thermodynamic definition and operating principle

A central formulation of the subject is the particle-exchange heat engine based on energy filtering. In the quantum-dot realization, particle-exchange engines use energy filtering to control a thermally driven particle flow between two heat reservoirs, do not require moving parts, and are therefore suitable for low-power applications and miniaturization. A quantum dot with resonance energy $\varepsilon$ connects hot and cold electronic reservoirs with temperatures $T_H$ and $T_C$ and chemical potentials $\mu_H$ and $\mu_C$. With a load resistance $R$, the steady-state circuit obeys $V=-IR$, and the electrical power delivered to the load is $P_{\rm th}=I^2R$. In the ideal limit of an infinitely sharp resonance, each transferred electron removes heat $Q_H=\varepsilon-\mu_H$ from the hot reservoir, deposits $Q_C=\varepsilon-\mu_C$ into the cold reservoir, and produces electrical work $T_H$0; the electronic efficiency is $T_H$1, where $T_H$2 is the electronic heat flow leaving the hot reservoir (Josefsson et al., 2017).

This formulation makes clear why particle-exchange engines are often steady-state rather than piston-like. In a continuous fermionic double-dot engine, particles are exchanged between hot and cold baths through a degenerate nanostructure, the power is $T_H$3, and the absorbed heat is $T_H$4, so that the efficiency is $T_H$5. In that setting, efficiency is fixed by the thermodynamic bias, and performance enhancement is identified with current enhancement. The relevant ideal benchmarks remain the Carnot efficiency $T_H$6 and the Curzon–Ahlborn efficiency $T_H$7 for finite-power operation (Um et al., 2021, Josefsson et al., 2017).

The nonequilibrium character of these machines is therefore not incidental. Their performance is controlled by the relation between charge or particle current, heat current, and the microscopic selectivity of the transport channel. Finite resonance width, higher-order tunneling, bath-induced decoherence, and load matching all determine how closely the device approaches the ideal one-energy transport limit.

2. Electronic energy-filtering realizations

The experimentally established solid-state archetype is the InAs/InP nanowire quantum-dot heat engine. In that device, thin InP segments define a small quantum dot between two electron reservoirs, the dot is tunnel-coupled with rate $T_H$8, and a gate electrode shifts the discrete resonance energy $T_H$9. The device is operated as a steady-state particle-exchange engine, and the electronic heat flow is obtained from a microscopic transport theory based on the real-time diagrammatic technique, which includes nonlinear transport and tunneling processes up to fourth order. The reported charging energy is $T_C$0 meV, the overall maximum power occurs for load resistances in the approximate range $T_C$1, the efficiency at the maximum-power operating point is in good agreement with the Curzon–Ahlborn prediction, and the overall maximum efficiency exceeds $T_C$2 while maintaining finite steady-state power output (Josefsson et al., 2017).

Strong electronic correlations substantially refine this picture. A non-equilibrium Green’s-function treatment of a strongly correlated single-level Anderson dot shows that Hartree–Fock is insufficient in the Coulomb-blockade regime, that the Hubbard-I approximation is appropriate for finite $T_C$3, and that a higher-order equation-of-motion decoupling beyond Hubbard-I is needed in the infinite-$T_C$4 limit. In the realistic load-driven engine, the operating point is fixed by the self-consistency condition $T_C$5, with output power $T_C$6. The analysis reproduces the optimal-load behavior observed experimentally: maximum power occurs around $T_C$7, whereas larger loads such as $T_C$8 and $T_C$9 push the efficiency to about $\mu_H$0 while still maintaining finite power (Verma et al., 2022).

The same thermodynamic logic has now been pushed to molecular scale. A single diradical molecule, SMe-2OS, bridging two gold electrodes in a three-terminal electromigrated break junction realizes a particle-exchange quantum heat engine that is autonomous, works without moving parts and without time-dependent driving, and is operated at cryogenic temperatures down to about $\mu_H$1 K. The measured thermocurrent maximum is about $\mu_H$2 pA, the total tunnel coupling is roughly $\mu_H$3 meV, and Kondo correlations with $\mu_H$4 K enhance both power and efficiency. The output power is reduced by about $\mu_H$5 when a magnetic field suppresses Kondo correlations, and the best reported normalized efficiency is $\mu_H$6 for $\mu_H$7 (Volosheniuk et al., 23 Aug 2025).

3. Atomic-collision and photon-exchange variants

Particle exchange need not be electronic. A direct atomic realization uses single Cesium impurity atoms immersed in an ultracold Rubidium bath. The Cesium working medium is the seven Zeeman sublevels of the $\mu_H$8 hyperfine ground manifold, and heat transfer occurs through inelastic spin-exchange collisions with spin-polarized Rubidium atoms. Each collision changes the Cs quasi-spin by $\mu_H$9 and the Rb bath atom by $\mu_C$0, so one engine atom and one bath atom exchange a single quantum of angular momentum and a corresponding single Zeeman energy quantum. The machine implements an endoreversible quantum Otto cycle with two collision-mediated heat strokes and two adiabatic magnetic-field ramps of $\mu_C$1 ms each. Heat exchange terminates after at most six spin-exchange steps per branch because the Cs quasi-spin has only seven levels. Experimentally, the measured efficiency is $\mu_C$2, the internal efficiency is $\mu_C$3, close to the predicted $\mu_C$4, and the maximal power output is $\mu_C$5 at a cycle time $\mu_C$6 (Bouton et al., 2020).

A photonic variant is developed in a four-level quantum system coupled to a unimodal cavity. The system has two lower nearly degenerate states $\mu_C$7 and $\mu_C$8 and two upper states $\mu_C$9 and $R$0; hot and cold baths drive population transfer through the upper manifold, while the transition $R$1 is coupled to a single cavity mode of frequency $R$2. In this model, work is identified with photon emission into the cavity mode, so the nonequilibrium fluctuations of interest are the photon-exchange statistics during the work-producing process. The authors formulate the dynamics through a full counting statistics approach combined with a quantum master equation, introducing a counting field $R$3 and a twisted Liouvillian $R$4. In the steady state, the dominant eigenvalue $R$5 is the cumulant generating function, and the first four cumulants $R$6, $R$7, $R$8, and $R$9 yield the mean photon current, variance, skewness, and kurtosis of the work channel (Sarmah et al., 2023).

These atomic and photonic examples broaden the scope of the term. What is exchanged need not be an electron traversing a quantum dot; it can be a counted atomic spin quantum or a counted cavity photon, provided the useful output is tied directly to the exchange process itself.

4. Coherence, interference, and fluctuation diagnostics

A recurrent theme in particle-exchange quantum heat engines is that coherence can be generated not by coherent tunneling in the system Hamiltonian but by interference in the system–bath coupling. In the degenerate double-quantum-dot engine, the relevant states are the empty state $V=-IR$0 and single-occupation states $V=-IR$1 and $V=-IR$2, with no coherent interdot tunneling. Coherence arises instead from interference between relaxation channels induced by the hot and cold fermionic baths. The steady-state particle current contains a classical part and a quantum part proportional to the coherence $V=-IR$3. In the symmetric coupling case, the steady state is generally classical and $V=-IR$4, except at the maximum-interference degenerate point $V=-IR$5, where a dark state appears and the long-time state becomes initial-condition dependent. When the symmetry of coupling configurations between dots and baths is broken, steady-state coherence survives and can enhance power in the nonlinear response regime; in linear response, the coherence-induced correction is non-positive (Um et al., 2021).

The cavity-coupled four-level engine turns this coherence problem around by treating it as an inverse inference task. The quantity of interest is the hot-bath-induced coherence between $V=-IR$6 and $V=-IR$7, parameterized by $V=-IR$8, where $V=-IR$9 denotes no hot-bath-induced coherence and $P_{\rm th}=I^2R$0 maximal alignment. The authors use the normalized cumulants $P_{\rm th}=I^2R$1 as input features and classify $P_{\rm th}=I^2R$2 into the four intervals $P_{\rm th}=I^2R$3, $P_{\rm th}=I^2R$4, $P_{\rm th}=I^2R$5, and $P_{\rm th}=I^2R$6. Among logistic regression, decision trees, support vector machines, random forests, AdaBoost, gradient boosting, and related models, K-nearest neighbors performs best overall, and the optimized KNN achieves validation accuracies above roughly $P_{\rm th}=I^2R$7 for the best mappings. The two-feature model using only the first two cumulants is the most efficient, indicating that the dominant predictive content lies in the mean and variance rather than in skewness and kurtosis. The final physical interpretation is that $P_{\rm th}=I^2R$8 or $P_{\rm th}=I^2R$9 favors the lowest coherence class, whereas $Q_H=\varepsilon-\mu_H$0 favors the highest coherence class (Sarmah et al., 2023).

This body of work shifts coherence from a formal attribute of the density matrix to an experimentally diagnosable transport signature. It also shows that, in particle-exchange engines, low-order fluctuation data may encode bath-induced coherence more directly than a complete microscopic reconstruction of dipole orientations or coupling amplitudes.

5. Autonomous conversion, collective symmetry, and generalized formulations

An autonomous extension couples a quantum-dot particle-exchange machine to a quantum mechanical resonator and asks whether electronic transport can be converted into mechanical self-oscillations. The model consists of a dot between two fermionic reservoirs and a resonator mode coupled through the dot occupation, with dynamics studied in the slow-transport regime $Q_H=\varepsilon-\mu_H$1 and for arbitrarily strong dot–resonator coupling $Q_H=\varepsilon-\mu_H$2. The central diagnostic is torotropy $Q_H=\varepsilon-\mu_H$3, introduced as a faithful measure of whether the resonator Husimi distribution is genuinely ring-shaped. In this framework, $Q_H=\varepsilon-\mu_H$4 if and only if self-oscillations are present. The electrical current through the dot can witness the onset of self-oscillations through a change in monotonicity, but under realistic conditions the self-oscillation regime lies entirely inside the heater region rather than the heat-engine region. The experimentally measurable conversion metric is $Q_H=\varepsilon-\mu_H$5, and strong dot–resonator coupling is found to be detrimental because Franck–Condon blockade suppresses the current that pumps the resonator (Sevitz et al., 22 Aug 2025).

Collective particle-exchange engines introduce a different generalization. An ensemble of $Q_H=\varepsilon-\mu_H$6 three-level systems collectively coupled to hot and cold baths and to a lasing transition can be decomposed into exchange-symmetry sectors using Schur–Weyl duality. The efficiency of emitted power into a dissipative load remains $Q_H=\varepsilon-\mu_H$7 for every irreducible representation, but ergotropy, lasing ergotropy, emitted power, spectral sharpness, and fluctuations depend strongly on the symmetry sector. Collective work extraction can extend beyond the temperature window of three-level lasing, yet in the lasing regime individual particles may outperform collective operation. The fully symmetric subspace often gives the largest ergotropy and emitted power, but bosonic symmetry is not always optimal, and some mixed-symmetry sectors can outperform others in specific parameter windows (Boeyens et al., 2024).

A more radical reformulation uses a $Q_H=\varepsilon-\mu_H$8-symmetric non-Hermitian two-level Hamiltonian to encode both working medium and reservoir action within a single operator. In that model, imaginary source/drain terms break ordinary probability conservation, so adiabatic evolution can change level populations through an imaginary geometric phase even when there are no inter-level transitions. The cycle is therefore described as an all-quantum-adiabatic-process cycle, interpreted as a particle-exchange engine because the non-Hermitian dynamics mimics exchange with reservoirs. For the rectangular cycle in parameter space, the efficiency is $Q_H=\varepsilon-\mu_H$9, identical in form to the Hermitian Otto result (Lin et al., 2015).

6. Conceptual boundaries and neighboring engine classes

The literature also makes clear what does not count as a particle-exchange quantum heat engine in the usual sense. The Josephson quantum heat engine combines thermoelectric transport in a normal-metal–ferromagnetic-insulator–superconductor tunnel junction with a Josephson weak link that converts a DC thermoelectric drive into an AC signal transferred contactlessly to an $Q_C=\varepsilon-\mu_C$0 load through electromagnetic induction. Its output power can reach about $Q_C=\varepsilon-\mu_C$1 pW into a $Q_C=\varepsilon-\mu_C$2 load under realistic conditions, but the device is best understood as a quantum thermoelectric heat engine with Josephson-mediated DC-to-AC conversion rather than as a particle-exchange engine (Marchegiani et al., 2016).

A different neighboring class is the interaction-driven many-particle quantum heat engine based on a one-dimensional Lieb–Liniger Bose gas. Here particle number is fixed throughout the cycle, there is no transfer of particles between reservoirs, and the relevant work strokes are produced by adiabatic changes of the interaction strength $Q_C=\varepsilon-\mu_C$3. In the low-temperature regime, the efficiency becomes universal, $Q_C=\varepsilon-\mu_C$4, and the average work per particle is maximal near the quantum critical region. The authors explicitly stress that this engine is not particle-exchange-based (Chen et al., 2018).

Closely related but still distinct are particle-statistics-driven engines. In hybrid Otto-like cycles built from a one-dimensional Lieb–Liniger gas, boson–fermion conversion strokes are inserted either into the adiabatic branches or into the thermal branches, and quantum statistics itself is treated as a thermodynamic resource. In the favorable T-engine configuration, the TG-BG cycle can reach Carnot efficiency in the degenerate regime. Since the particle number remains conserved and the key resource is the conversion between bosonic and fermionic thermodynamic behavior, this is best viewed as an adjacent extension of the particle-exchange idea rather than as the standard reservoir-to-reservoir carrier-transfer picture (Menon et al., 25 Mar 2025).

Taken together, these distinctions delimit the field. In the narrow sense, a particle-exchange quantum heat engine converts heat to work by selective transfer of discrete quanta across an energy-selective channel. In the broader contemporary literature, the same logic has been generalized to counted atomic collisions, cavity-photon work channels, collective exchange-symmetry sectors, autonomous self-oscillation converters, and effective non-Hermitian reservoir models. The unifying question is how quantum selectivity, coherence, correlations, and symmetry alter the conversion of exchanged quanta into useful work.