Two-Stroke Quantum Refrigerator

• Two-stroke quantum refrigerators are quantum thermal machines that alternate between unitary evolution and thermal reset steps for efficient heat transfer.
• They employ two-level systems and optimized cycle conditions to achieve refrigeration while adhering to thermodynamic and finite-time constraints.
• Catalytic enhancements and advanced control protocols extend operational regimes, boosting performance metrics like COP in superconducting and NMR platforms.

A two-stroke quantum refrigerator is a quantum thermal machine that transfers heat from a cold reservoir to a hot reservoir through a cycle composed of two alternating stages: a driven or unitary stroke that redistributes populations or performs work, and a thermalization or reset stroke that restores the relevant bath-dependent marginals. In recent arXiv literature, the term covers both literal two-step machines—such as two-level-system devices operated by unitary exchange followed by bath re-equilibration—and reduced descriptions of Otto-like refrigerators whose dynamics are optimized as a cooling/work stroke followed by a reset stroke (Bhattacharjee et al., 2019, Fu et al., 16 Jul 2025, Menczel et al., 2019). A recurrent terminological point is that not every compact quantum refrigerator is two-stroke in the strict sense: the reversed-coupling two-qubit absorption fridge, for example, is explicitly a four-stroke autonomous absorption refrigerator rather than a two-stroke machine (Silva et al., 2015).

1. Cycle architecture and state preparation

In its most direct form, a two-stroke quantum refrigerator uses a finite-dimensional working medium coupled to two reservoirs at different temperatures. The basic cycle consists of a first stroke in which the working subsystems are decoupled from the baths and undergo a controlled unitary evolution, followed by a second stroke in which the interaction is switched off and the subsystems re-equilibrate with their respective baths. In the two-level magnetometer of Hofer and Brask, the working substance is a pair of two-level systems (TLSs): one TLS with splitting $2un$ generated by a weak unknown magnetic field, and a second TLS with splitting $2k$ generated by a known tunable field. The former is initially thermalized with the cold bath at temperature $c$, the latter with the hot bath at temperature $h$, and heat exchange occurs only during the thermalization stroke (Bhattacharjee et al., 2019).

A closely related literal two-stroke refrigerator is analyzed in the catalytic framework of 2025. There the working medium again consists of two TLSs, one associated with the hot side and one with the cold side, but the cycle is assisted by an auxiliary catalyst of dimension $d$. The initial state is a product $\rho=\tau_h\otimes\tau_c\otimes\rho_s$, a global unitary $U$ acts during the work stroke, and the catalyst is required to be cyclic in the sense that $\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$. The second stroke detaches the catalyst and rethermalizes the hot and cold TLSs with their reservoirs, thereby restoring the initial product structure (Fu et al., 16 Jul 2025).

A more abstract formulation is given in the mesoscopic optimal-control literature. There the refrigerator consists of a working system $S$, a cold reservoir at temperature $T_c$, a hot reservoir at $2k$0, and an externally controlled parameter $2k$1 that both drives the system and switches its reservoir coupling. The cycle is split by a threshold value $2k$2: during the work stroke $2k$3, one has $2k$4 and the system is coupled to the cold reservoir; at $2k$5, the control is jumped instantaneously above threshold; during the reset stroke $2k$6, one has $2k$7 and the system releases heat to the hot reservoir (Menczel et al., 2019).

2. Thermodynamic description and mode classification

The thermodynamic bookkeeping of two-stroke quantum refrigerators depends on convention, but the central observables are always the cold-side heat extraction, the hot-side heat release, and the required work input. In the two-stroke thermal machine of Hofer and Brask, $2k$8 and $2k$9 denote the heat extracted from the hot and cold baths, respectively, and since the cycle closes the net work is

$c$0

With that sign convention, the machine is a refrigerator when

$c$1

an engine when $c$2, and an accelerator when $c$3. The reported two-stroke machines realize the first three modes, and the refrigerator–engine transition is the operationally important one because the heat currents reverse sign there (Bhattacharjee et al., 2019).

In the catalytic refrigerator literature, the definitions are phrased in the conventional refrigerator form. The heat released to the hot bath and absorbed from the cold bath are

$c$4

and the work input is

$c$5

The coefficient of performance is

$c$6

with refrigeration requiring $c$7, $c$8, and $c$9. The second law yields

$h$0

and hence the Carnot bound

$h$1

The same work also states a universal ordering across operational modes,

$h$2

This ordering makes explicit that refrigeration occupies an intermediate thermodynamic sector rather than a sharply isolated one (Fu et al., 16 Jul 2025).

At the level of optimization theory, efficiency is commonly written as

$h$3

with the Carnot upper bound

$h$4

A persistent conclusion is that maximum efficiency occurs in the reversible limit but with vanishing cooling power, whereas protocols designed for strong finite-time cooling necessarily operate away from that limit (Menczel et al., 2019).

3. Two-level working media, transition points, and the non-catalytic limit

The simplest explicit two-stroke quantum refrigerator is built from two TLSs. In the magnetic-field-sensing thermal machine, the refrigerator regime is controlled by the known splitting $h$5. The central transition occurs when the excited-state populations match,

$h$6

with

$h$7

This gives the mode-switching condition

$h$8

At that point,

$h$9

For the swap two-stroke thermal machine, the heat and work are

$d$0

Hence $d$1 corresponds to refrigeration, $d$2 first gives heat-engine operation near the transition and then accelerator behavior farther away. For the more general mixing interaction

$d$3

all three thermodynamic quantities acquire a common factor $d$4, so the transition condition remains $d$5 independent of the entanglement generated during the unitary stroke (Bhattacharjee et al., 2019).

In the non-catalytic two-TLS refrigerator of 2025, the work stroke is a global unitary on the hot and cold TLSs. Because $d$6 depends only on diagonal populations, the dephased post-unitary state can be represented by a bistochastic map, and Birkhoff’s theorem then implies that the optimal cooling is attained by a specific permutation. Without a catalyst, the Hilbert-space dimension is $d$7, so there are $d$8 permutations; only four of them yield $d$9. For the optimal permutation,

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$0

and with

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$1

one obtains

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$2

The corresponding optimal coefficient of performance is

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$3

which is exactly the Otto COP. Refrigeration in this non-catalytic setting requires

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$4

This establishes the standard two-TLS refrigerator as a bound-limited baseline: optimal within its restricted permutation class, but constrained to the Otto form and to a restricted frequency–temperature region (Fu et al., 16 Jul 2025).

4. Catalytic enhancement and self-contained variants

Catalytic two-stroke refrigeration augments the two-TLS working medium by an auxiliary system that is returned to its original marginal state after each cycle. The catalyst is not an energy sink or source in the thermodynamic accounting of the cycle; rather, it enlarges the space of admissible population permutations. The 2025 catalytic construction enforces exact cyclicity by balancing the net population flows through the catalyst levels,

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$5

A first catalytic permutation produces a net outflow $\rho=\tau_h\otimes\tau_c\otimes\rho_s$6 from the cold subspace and net inflow $\rho=\tau_h\otimes\tau_c\otimes\rho_s$7 to the hot subspace, leading to

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$8

A second construction yields

$\rho=\tau_h\otimes\tau_c\otimes\rho_s$9

In this framework, catalysis has two precisely stated consequences: it can make the COP exceed the Otto bound, and it can enlarge the operational frequency–temperature region so that refrigeration becomes possible in regimes inaccessible without a catalyst. The paper further states that

$U$0

guarantees that the COP does not exceed Carnot, while the expanded-regime cooling condition is

$U$1

In the limit $U$2, the cooling condition simplifies essentially to $U$3, and refrigeration may remain possible even when $U$4 (Fu et al., 16 Jul 2025).

A distinct but conceptually adjacent line of work studies self-contained minimal refrigerators. The 2024 NMR experiment realizes the smallest self-contained refrigerator using three effective two-level systems, with the key degeneracy condition

$U$5

The cooling transition is the exchange

$U$6

generated by

$U$7

The necessary and sufficient cooling condition is

$U$8

the coefficient of performance is

$U$9

and it obeys the Carnot-like bound

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$0

The paper does not present this machine as a textbook two-stroke Otto refrigerator; rather, it describes a three-qubit self-contained exchange that can be viewed effectively as a two-stage cycle of exchange evolution and ancilla reset. This suggests that the boundary between “two-stroke” and “self-contained exchange” refrigerators is partly architectural rather than purely thermodynamic (Huang et al., 2024).

5. Otto decomposition, finite-time control, and coherence effects

A substantial part of the literature treats four-stroke Otto refrigerators through an effective two-stroke decomposition. In the universal optimization scheme for mesoscopic refrigerators, the state vector $\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$1 obeys

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$2

the cold heat extracted during the work stroke is

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$3

and the hot-side release during the reset stroke is

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$4

The control problem is solved by combining stochastic or quantum thermodynamics with Pontryagin’s minimum principle and a three-step procedure: optimize the work stroke for fixed initial conditions, optimize the reset stroke to restore the initial state, and then optimize the initial conditions themselves. In the fast-driving regime, any protocol is effectively replaced by the step protocol

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$5

whereas in the slow-driving regime the state remains close to instantaneous equilibrium and the reset stroke becomes negligible at leading order (Menczel et al., 2019).

The shortcut-to-adiabaticity Otto refrigerator makes this reduction especially explicit. Its full cycle contains two unitary strokes and two isochores, but the isochoric times are taken to be negligible compared with the unitary strokes, so that

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$6

for equal stroke durations. The local counterdiabatic protocol uses the effective oscillator Hamiltonian

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$7

with

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$8

and defines a superadiabatic COP that includes the cost of the control field. The reported hierarchy is

$\operatorname{Tr}_{h,c}[U\rho U^\dagger]=\rho_s$9

for a broad range of stroke times, with the shortcut-enhanced refrigerator outperforming the conventional finite-time device except at very short cycle times. The same work derives quantum-speed-limit upper bounds on COP and cooling power that are tighter than the second-law bound (Abah et al., 2019).

In superconducting-qubit Otto refrigerators, finite-time behavior introduces additional structure. The 2016 analysis of a qubit coupled to two LC resonators identifies a nearly adiabatic low-frequency regime in which cooling power scales as $S$0, an intermediate regime approximating the ideal Otto cycle, and a high-frequency coherent regime with oscillatory power. A central conclusion is that quantum coherent effects decrease both cooling power and coefficient of performance relative to purely classical dynamics, while truncated trapezoidal driving yields higher cooling power and efficiency than a standard sinusoidal drive among the waveforms considered (Karimi et al., 2016). A related two-qubit Otto study shows that correlated or anticorrelated noise can create protected states at $S$1, induce very long relaxation times near those values, and prevent the expected linear scaling of cooling power with the number of qubits even in the absence of direct qubit–qubit interaction (Karimi et al., 2017).

6. Experimental platforms and application-oriented realizations

Superconducting circuits are a leading platform for two-stroke or effectively two-stroke refrigeration. A Cooper-pair-box charge qubit capacitively coupled to two $S$2 coplanar-waveguide resonators has been fabricated and spectroscopically characterized as a refrigerator architecture. The measured resonator frequencies are

$S$3

with extracted effective couplings

$S$4

By driving the offset charge $S$5, the qubit can be swept between resonance with the two resonators; after replacing the resonator terminations by normal-metal resistors, this architecture implements a quantum Otto refrigerator with cooling condition

$S$6

For realistic parameters, including low resonator quality factors $S$7, $S$8, and drive shape parameter $S$9, simulations give peak cooling power around $T_c$0 in the idealized case and about $T_c$1 when quasiparticle poisoning is included; the paper states that this remains measurable with NIS thermometry (Guthrie et al., 2021).

The self-contained NMR refrigerator provides the first explicit experimental realization of minimal quantum refrigeration with no net external work. Using a Bruker 300 MHz NMR spectrometer and three selected $T_c$2C nuclei in crotonic acid dissolved in $T_c$3-acetone, the experiment decomposes the exchange evolution into forty steps of one- and two-qubit gates, verifies that the total work sums to zero, and observes cyclic cooling of the target spin. Under repeated reset of the auxiliary spins, the target temperature converges to the reported bound

$T_c$4

Although the paper does not cast the device as a standard two-stroke engine, its exchange-plus-reset structure places it in the immediate conceptual neighborhood of two-stroke quantum refrigeration (Huang et al., 2024).

Application-driven variants often adopt a two-stroke interpretation even when their primary goal is active qubit reset rather than heat-engine analysis. The quantum-circuit refrigerator based on a voltage-controlled SINIS structure is not explicitly presented as a standard two-stroke refrigerator, but its operation has a clear off-stroke/on-stroke structure: a low-bias off state in which QCR-induced transitions are negligible and a high-bias on state in which the device acts as a strong cold bath. For typical experimental parameters, the reported reset performance reaches fidelities above $T_c$5 on nanosecond timescales: about $T_c$6 for a transmon in the low-temperature regime and about $T_c$7 for a capacitively shunted flux qubit (Hsu et al., 2020). A separate qubit-cooling study analyzes two operationally two-stroke refrigerators: one with a driven auxiliary qubit and repeated reset, and a second with two auxiliary qubits and no frequent reset. There the central result is that the no-reset scheme has the larger coefficient of performance, and for $T_c$8 it also removes more heat from the target qubit than the repeated-reset scheme (Okane et al., 2022).

Taken together, these results show that the two-stroke quantum refrigerator is not a single fixed model but a family of closely related architectures united by the alternation of a population-redistribution stage and a reset or thermalization stage. Within that family, the principal research directions are the characterization of transition conditions and sign reversals, the extension of the accessible cooling region by catalysts or auxiliary systems, and the optimization of finite-time control under coherence, noise, and hardware constraints (Bhattacharjee et al., 2019, Fu et al., 16 Jul 2025, Menczel et al., 2019).