Quantum Thermal Logic Gates

• Quantum thermal logic gates are devices that encode logical information using thermal variables like temperature and heat current, establishing Boolean operations via thermodynamic thresholds.
• They integrate diverse methodologies—from superconducting circuits and quantum dots to spintronic platforms—demonstrating multiple frameworks for thermal logic within quantum thermodynamics.
• These gates leverage both direct heat-based signals and resilience against thermal noise, paving the way for fault-tolerant quantum computation and innovative thermodynamic circuit designs.

Quantum thermal logic gates are gate primitives and gate frameworks in which thermal variables enter the logic operation at a fundamental level. In the literature, this designation spans several related constructs: Boolean gates whose inputs and outputs are temperatures or heat currents; Gibbs-preserving two-level maps that function as elementary thermodynamic gates; ordinary quantum gates analyzed through non-equilibrium thermodynamics; and quantum-information gates designed to remain operative when the underlying systems are hot or explicitly thermalized (Paolucci et al., 2017, Lostaglio et al., 2016, Cimini et al., 2020, Petit et al., 2019, Ghosh et al., 4 Jun 2026). The field therefore lies at the intersection of quantum thermodynamics, mesoscopic heat transport, superconducting and spintronic device physics, and fault-tolerant quantum control under thermal constraints.

1. Terminological scope and conceptual taxonomy

Taken together, the published record indicates that “quantum thermal logic gate” does not denote a single architecture. In one usage, logical states are encoded directly in thermal observables. The phase-tunable Josephson proposal assigns each logic variable a local electronic temperature $T_X$, with “cold” representing logic 0 and “hot” representing logic 1, and realizes NOT, AND, and OR by steering heat through phase-coherent superconducting valves (Paolucci et al., 2017). Closely related coupled-quantum-dot and superconducting-transistor proposals likewise use source temperatures or gate temperatures as inputs and decode the output from a heat current crossing a threshold (Majland et al., 2019, Ghosh et al., 4 Jun 2026).

A second usage is resource-theoretic. Here the basic thermal “gate” is an elementary thermal operation: a completely-positive, trace-preserving map acting non-trivially only on two energy eigenstates while preserving the Gibbs state at inverse temperature $\beta$. In this formulation, partial thermalisations and $\beta$-swaps play the role of elementary building blocks for thermodynamic circuits (Lostaglio et al., 2016).

A third usage concerns the thermodynamic characterization of quantum gates that are not themselves heat-based Boolean elements. The experimental optical study of a two-qubit controlled-unitary gate reconstructs the statistics of energy and entropy fluctuations generated by the gate, connects these to gate performance, and tests Landauer-type bounds at the single-quantum level (Cimini et al., 2020). In this sense, a “quantum thermal logic gate” is an ordinary logic primitive analyzed through the lens of non-equilibrium quantum thermodynamics.

A fourth usage appears in quantum-information architectures that operate away from the mechanical ground state. Thermofield Dynamics treats a thermalized logic gate as the thermal counterpart of an ordinary unitary, while silicon spin qubits above one Kelvin and logically encoded hot-system gates show that universal or high-fidelity quantum logic can persist under substantial thermal occupation (Trindade et al., 2012, Petit et al., 2019, Riera-Sàbat et al., 2023). A common misconception is therefore that the field is only about “computing with heat.” The broader literature includes both computation by heat flow and quantum computation under thermal constraints.

2. Information carriers and hardware realizations

The physical carrier of the logical variable differs substantially across platforms. Some architectures use temperature itself as the bit value, some use heat current as the output signal, and some use thermally induced modifications of an otherwise conventional quantum gate.

Platform	Logical variable	Physical realization
Phase-tunable Josephson logic	Local electronic temperature $T_X$	SQUIPT valve + N-FI-S actuator (Paolucci et al., 2017)
Superconducting thermal transistor logic	Gate temperature $T_G$, drain heat current $J_D$	Qubit–qutrit cQED device (Majland et al., 2019)
Coupled-quantum-dot logic	Source temperatures, output $|J_Q|$	Two single-level quantum dots with Coulomb repulsion (Ghosh et al., 4 Jun 2026)
Autonomous thermal machines	Bath inverse temperatures $\beta_k$, output $\beta_z^\infty$	Few-qubit “thermodynamic neuron” (Lipka-Bartosik et al., 2023)
PT-symmetric spintronic logic	Current-controlled gain/loss, output temperature rise $T_{\rm out}$	Coupled YIG waveguides with Pt spacer (Wang et al., 2023)
Thermalized or hot-system quantum gates	Gate fidelity under thermal occupation	Bosonic-mode TFD encoding; silicon spin qubits; logical hot-system encoding (Trindade et al., 2012, Petit et al., 2019, Riera-Sàbat et al., 2023)

In the superconducting phase-tunable architecture, the proof-of-principle Al/Cu implementation uses $\beta$0 and $\beta$1, with tolerance windows $\beta$2 and $\beta$3. A power-supply reservoir at $\beta$4 feeds a tunable thermal valve, the output node floats to a steady temperature $\beta$5, and input nodes act as actuators through inductively coupled coils. The valve is a temperature-biased superconducting quantum interference proximity transistor, and the actuator is an N-FI-S tunnel junction generating a thermocurrent $\beta$6 that controls the valve flux (Paolucci et al., 2017).

In the superconducting quantum thermal transistor, Majland et al. use a three-terminal circuit-QED setup consisting of a qubit, a qutrit, and a mediating resonator. Source, drain, and gate baths are resistor-based circuits tuned to specific transition frequencies, and the logical state is extracted from the steady-state source or drain heat current after thresholding (Majland et al., 2019).

The coupled-quantum-dot proposal of 2026 uses two single-level quantum dots $\beta$7 and $\beta$8 with strong inter-dot Coulomb repulsion $\beta$9, each tunnel-coupled to metallic reservoirs. Inputs are encoded on source leads $\beta$0 as logic-0 for $\beta$1 and logic-1 for $\beta$2. The output is decoded from $\beta$3 using thresholds $\beta$4 and $\beta$5 (Ghosh et al., 4 Jun 2026).

The autonomous-machine approach does not transmit logic through a wire-like channel. Instead, it sets the temperatures of several baths according to the input, lets a few-qubit machine relax to a non-equilibrium steady state, and reads the computation from the temperature of an auxiliary finite-size reservoir. This defines a “thermodynamic neuron” rather than a gate in the circuit-theory sense (Lipka-Bartosik et al., 2023).

3. Thermodynamic and open-system formalisms

Despite their hardware diversity, these systems are described by a small number of recurring formalisms: Gibbs-preserving maps, Lindblad master equations, steady-state heat-balance equations, and fluctuation relations.

For elementary thermal operations, the defining object is a two-level Gibbs-stochastic matrix acting on populations of two energy eigenstates $\beta$6. If $\beta$7, the map is parameterized by $\beta$8 and takes the form

$\beta$9

while all other levels are unchanged. A full $T_X$0-swap corresponds to $T_X$1. This provides the cleanest abstract definition of a thermal gate as a local Gibbs-preserving population transformation (Lostaglio et al., 2016).

In the phase-tunable Josephson architecture, the output temperature is fixed by the steady-state balance equation

$T_X$2

The valve heat current is controlled by the phase-dependent density of states of the proximized wire, and the heat flow can also be expressed as

$T_X$3

By choosing $T_X$4, the minigap nearly blocks $T_X$5; at $T_X$6, the transmitted heat is maximized (Paolucci et al., 2017).

The cQED thermal transistor is formulated with an effective six-level Hamiltonian

$T_X$7

coupled to three baths through a Born-Markov-Lindblad description. The steady-state heat current into bath $T_X$8 is

$T_X$9

and the amplification factor is defined as

$T_G$0

This imports transistor concepts into a thermal quantum device (Majland et al., 2019).

The coupled-quantum-dot architecture employs

$T_G$1

with a Lindblad master equation for the reduced density matrix. At steady state, the heat current entering from lead $T_G$2 is

$T_G$3

The logic function is therefore encoded in a transport observable rather than in a state occupation or a unitary truth table (Ghosh et al., 4 Jun 2026).

The experimental energetics of a two-qubit controlled-unitary gate use the two-point measurement scheme on local Hamiltonians and define the stochastic energy change as $T_G$4. For an initial thermal state of qubit $T_G$5, the experiment recovers

$T_G$6

This places quantum logic energetics within the same fluctuation-theorem structure as other non-equilibrium thermal processes (Cimini et al., 2020).

4. Realizations of Boolean gate families

The first complete gate family in this area was the phase-tunable thermal logic proposal based on superconducting proximity circuits. The NOT gate uses one normally open valve and two coils, one providing a fixed opening flux $T_G$7 and the other supplying an input-dependent flux $T_G$8. The design chooses the mutual inductance so that $T_G$9. Then $J_D$0 yields $J_D$1, the valve remains open, $J_D$2, and the output is 1; $J_D$3 yields $J_D$4, the valve closes, $J_D$5, and the output is 0. The same platform realizes AND with one valve and two input coils, and OR with two parallel SQUIPTs sharing an output node. In all cases the output temperature is obtained by solving the steady-state heat-balance equation, and input fluctuations of $J_D$6 still leave the output within the correct logic window (Paolucci et al., 2017).

Majland et al. show that a quantum thermal transistor in superconducting circuits can be assembled into NOT, AND, and OR gates. NOT uses a single transistor with the gate temperature as input and an inverted threshold rule on the drain current. AND is obtained by placing two transistors in series, so heat reaches the final drain only when both transistors conduct. OR is obtained by placing two transistors in parallel and summing the branch drain currents. This construction is significant because it translates Boolean circuit composition almost literally into series and parallel compositions of thermal conductors (Majland et al., 2019).

The coupled-quantum-dot proposal pursues the classical-circuit analogy even more explicitly. By appropriate coupling of leads, it realizes Buffer, NOT, OR, AND, NOR, and NAND. The Buffer uses a single source $J_D$7 and drain $J_D$8, with a “clockwise” cycle $J_D$9 producing $|J_Q|$0, so that $|J_Q|$1 and $|J_Q|$2. NOT adds a hot invert lead $|J_Q|$3 and reads out $|J_Q|$4; OR uses two parallel sources $|J_Q|$5; AND adds a hot control lead $|J_Q|$6 to the OR geometry; and NOR and NAND are obtained by inversion. The paper states a one-to-one correspondence with the structure of classical electronic logic-gate circuits (Ghosh et al., 4 Jun 2026).

In the PT-symmetric spintronic proposal, the logical signal is embodied in magnetic excitations transported through laterally coupled YIG waveguides separated by Pt. Balanced gain and loss are introduced through spin-orbit torque, leading to a PT-symmetric effective two-mode matrix and an exceptional point when $|J_Q|$7. The device supports a thermal diode and logic functions: OR uses microwave inputs on the two waveguides and reads $|J_Q|$8 in WG1; AND uses the current state together with a microwave input on WG2; NOT maps the current state to the opposite logic value at the output temperature. Here the logic mechanism is not static thermal thresholding alone but current-tuned non-reciprocal magnon-heat transport near the exceptional point (Wang et al., 2023).

Autonomous quantum thermal machines generalize the gate notion further. A three-qubit collector with an energy-preserving interaction and a modulator implements NOT through the virtual temperature

$|J_Q|$9

with the output inferred from the steady-state $\beta_k$0 of a finite reservoir. The same framework explicitly realizes NOR and 3-MAJORITY, and networks of such “thermodynamic neurons” can realize non-linearly-separable functions such as XOR by layering linearly separable modules (Lipka-Bartosik et al., 2023). This suggests a shift from gate-based thermal logic to thermodynamic neural computation.

A closely related but distinct area is near-field radiative thermal logic in nanoparticle networks. Kathmann et al. realize NOT, UNIT, OR, NOR, AND, and NAND using VO$\beta_k$1, SiC, and SiO$\beta_k$2 nanoparticles, while emphasizing that many-body non-additivity prevents naive cascading of gates (Kathmann et al., 2020). Although this work is not formulated as a quantum-dot gate platform, it is an adjacent thermal-logic framework and is explicitly noted as potentially extendable to quantum dots.

5. Energetics, coherence, and the status of Landauer-type bounds

The energetics of quantum logic operations has been placed on an experimental footing by the optical implementation of a controlled-unitary two-qubit gate. The gate is generated by a time-independent Hamiltonian

$\beta_k$3

with

$\beta_k$4

and is implemented in linear optics using photon polarizations $\beta_k$5, $\beta_k$6. The experiment reconstructs the full joint probability table $\beta_k$7 via two-point measurements and shows that the first five moments of $\beta_k$8 oscillate periodically, peaking at $\beta_k$9. The energy distribution is positively skewed with a fat right tail, while the entropy distribution is symmetric and nearly bell-shaped at the same times (Cimini et al., 2020).

The same experiment provides a direct single-shot, single-quantum verification of the non-equilibrium Clausius-Landauer relation

$\beta_z^\infty$0

For an initial thermal state of qubit $\beta_z^\infty$1 with $\beta_z^\infty$2, the measured ratio $\beta_z^\infty$3 always lies above $\beta_z^\infty$4, and the bound is tightest when coherence generation is minimal. An important design implication reported in the paper is that entropy production is tightly linked to dynamically generated quantum coherence $\beta_z^\infty$5; operating at times or control angles where coherence creation is minimal reduces $\beta_z^\infty$6 and therefore the energetic cost (Cimini et al., 2020).

At the more abstract level of resource theory, elementary thermal operations clarify that thermal gate sets are constrained in a way unlike ordinary universal gate libraries. At infinite temperature, Gibbs-stochastic matrices reduce to doubly stochastic matrices, and two classical theorems imply that arbitrary transformations can be decomposed into two-level $\beta_z^\infty$7-transforms. At finite $\beta_z^\infty$8, however, both the Muirhead–Hardy–Littlewood–Polya and Birkhoff theorems fail in the Gibbs-preserving setting. In particular, there exist three-level transitions allowed by full Thermal Operations but impossible by any finite sequence or mixture of two-level elementary thermal operations. The support size of the population vector becomes an ETO monotone. Thus, thermal gate universality is temperature-dependent and, in general, strictly weaker than full Thermal Operations (Lostaglio et al., 2016).

This is a central conceptual distinction. A common oversimplification is to equate “elementary thermal gates” with a universal thermodynamic gate set. The finite-temperature resource theory shows that such a gate set can be physically motivated and operationally useful while still failing to generate all Gibbs-preserving transitions available to more general Thermal Operations (Lostaglio et al., 2016).

6. Quantum logic under explicit thermal occupation

Quantum thermal logic also encompasses ordinary quantum-information gates subjected to thermalization or executed between hot subsystems. In Thermofield Dynamics, the thermal counterpart of a unitary $\beta_z^\infty$9 is

$T_{\rm out}$0

and its action on a thermalized input obeys

$T_{\rm out}$1

Applied to a bosonic-mode encoding of a CNOT, this framework shows that temperature acts as quantum noise, transforming pure states into statistical mixtures. The fidelity decreases with temperature; the Mandel parameter crosses from sub-Poissonian to Poissonian to super-Poissonian behavior at a critical thermal occupation $T_{\rm out}$2; and the negative regions of the Wigner function shrink as $T_{\rm out}$3 increases (Trindade et al., 2012).

A more hardware-oriented realization is provided by Petit et al., who demonstrated universal single- and two-qubit logic above one Kelvin in silicon quantum dots. Single-qubit control is performed by electron-spin resonance with $T_{\rm out}$4 and $T_{\rm out}$5, $T_{\rm out}$6-pulse durations $T_{\rm out}$7, and randomized-benchmarking fidelities $T_{\rm out}$8 and $T_{\rm out}$9. The exchange interaction $\beta$00 is tuned from $\beta$01 up to $\beta$02, and coherent two-qubit controlled rotations are demonstrated at $\beta$03. The measured $\beta$04 values, $\beta$05 for Q1 and $\beta$06 for Q2 at weak exchange, change little between $\beta$07 and $\beta$08, while $\beta$09 up to $\beta$10 had been shown previously. This establishes that universal gate sets can survive in a regime far hotter than the standard dilution-refrigerator baseline (Petit et al., 2019).

Riera-Sàbat, Sekatski, and Dür address a different thermal problem: unknown positions or positional fluctuations of hot physical qubits. They encode each logical qubit into $\beta$11 physical two-level systems so that the effective logical coupling

$\beta$12

becomes flatter as a function of the noisy coordinates. The gate fidelity for classical positional noise is

$\beta$13

with an analogous quantum expression in the mechanical eigenbasis. Numerical optimization indicates that $\beta$14 already yields order-of-magnitude error suppression for Gaussian fluctuations of width $\beta$15 a few lattice spacings, and the paper reports significant improvement of gate fidelities by enlarging the logical system (Riera-Sàbat et al., 2023).

These works collectively show that “thermal” in quantum logic need not mean that heat current is the logic signal. It can also mean that the gate model explicitly includes thermal occupation, thermal noise, or hot-device operation.

7. Performance metrics, cascading, and design constraints

Reported performance metrics vary widely because the underlying devices are physically distinct. In the Josephson phase-tunable proposal, the switching speed is limited by electron–phonon relaxation in the output lead, with $\beta$16 and $\beta$17 at $\beta$18, while higher-$\beta$19 superconductors can raise operation to $\beta$20. The superconducting actuation loop is described as dissipationless, the only lost power being electron–phonon heat leakage in the valve, and the error rate is stated to be negligible $\beta$21 at $\beta$22. Fan-out is $\beta$23 at $\beta$24, with a phase-tunable temperature amplifier proposed for higher-temperature cascading (Paolucci et al., 2017).

In the cQED thermal transistor, the switching time is estimated as $\beta$25, set by the smallest nonzero Lindblad rate or the internal coupling $\beta$26. Off-state leakage is suppressed by $\beta$27 band-pass filtering, and the thermal amplification ratio is $\beta$28, improving with anharmonicity $\beta$29 (Majland et al., 2019).

The coupled-quantum-dot proposal quotes $\beta$30, a typical detection noise $\beta$31, measured heat-current sensitivity down to $\beta$32 in superconducting nano-calorimeters, and a typical gate error rate $\beta$33. Its stated scalability challenges are thermal crosstalk, gate-to-gate variability of dot levels and tunnel rates, and the need for on-chip thermal isolation and interconnect (Ghosh et al., 4 Jun 2026).

For PT-symmetric spintronic gates, the exceptional-point current density is

$\beta$34

which the paper states is well below typical auto-oscillation thresholds in YIG|Pt. The use of YIG with $\beta$35 supports long-range magnon transport, while balanced gain and loss produce strong enhancement near the exceptional point (Wang et al., 2023).

Cascading is one of the main architectural fault lines in the field. In the superconducting phase-tunable scheme, input and output use the same variable—temperature—and are galvanically decoupled by inductive coupling, so gates can be cascaded directly (Paolucci et al., 2017). By contrast, near-field radiative logic is intrinsically non-additive: Kathmann et al. show that an OR gate followed by a NOT gate does not automatically yield a NOR gate unless the entire geometry is redesigned as a single many-body problem (Kathmann et al., 2020). The autonomous thermal-machine approach resolves composability differently, by feeding the steady-state temperature of one finite reservoir into the next layer, effectively turning gate composition into network design (Lipka-Bartosik et al., 2023).

The present state of the subject therefore suggests two broad design logics. One is circuit-like: series and parallel composition, thresholding, and direct gate libraries. The other is thermodynamic: steady-state transduction, virtual temperatures, and network composition. Both are constrained by thermal noise, dissipation, and crosstalk, but they differ sharply in whether they emulate conventional digital circuits or exploit genuinely non-circuit thermodynamic dynamics.