Quantum Szilard Engines

Updated 16 March 2026

Quantum Szilard engines are thermodynamic devices that convert information from quantum measurements into work by exploiting quantum mechanics and statistical physics.
They operate through a four-stroke cycle—adiabatic insertion, measurement, expansion, and barrier removal—where quantum information directs net energy conversion.
They provide a platform to probe fundamental thermodynamic laws, validate fluctuation theorems, and benchmark the conversion of information into thermodynamic resources.

A quantum Szilard engine (QSZE) is a thermodynamic device that exploits the interplay of quantum mechanics, statistical physics, and information theory to convert thermal energy into work by extracting and utilizing information about a quantum system. Generalizing the classical Szilard engine thought experiment to the quantum domain requires precise modeling of measurement-induced state changes, transformation of quantum distributions, particle indistinguishability and statistics, and the incorporation of quantum information-theoretic quantities into the thermodynamic bookkeeping. The QSZE not only forms a testbed for foundational questions about the second law and Maxwell’s demon at the quantum scale, but also benchmarks the operational equivalence between information and thermodynamic resources.

1. Canonical Model and Thermodynamic Cycle

The prototypical QSZE consists of a single particle of mass $m$ confined in a one-dimensional infinite square well of width $L$ , coupled to a heat bath at temperature $T$ (Li et al., 2012). The system Hamiltonian is

$H = -\frac{\hbar^2}{2m}\frac{d^2}{dx^2}$

with eigenstates $|E_n\rangle$ and eigenvalues $E_n(L)=\hbar^2\pi^2 n^2 / (2m L^2)$ for $n=1,2,\dots$ . The initial equilibrium state is characterized by the density matrix $\rho_0 = \sum_n P_n |E_n\rangle\langle E_n|$ where $P_n = e^{-\beta E_n}/Z(L)$ , $Z(L) = \sum_n e^{-\beta E_n}$ , $\beta = (k_BT)^{-1}$ .

A complete QSZE cycle comprises four strokes:

Adiabatic Insertion: An infinite potential barrier is slowly inserted at $x=L/2$ , splitting the system into two wells of width $L/2$ . Quantum mechanically, odd and even parity eigenstates respond differently, resulting in level shifts and a net insertion work

$W_1 = \sum_{k=1}^\infty P_{2k-1}(L)\bigl(E_{2k}(L) - E_{2k-1}(L)\bigr).$

Measurement: A projective measurement localizes the particle into one side (say, left), producing a state $\rho^{(L)}$ , erasing classical entropy $S_c = k_B\ln2$ , but costing no energy ( $Q_2=W_2=0$ ).
Expansion: The barrier is allowed to move from $x=L/2$ $x = L /2$ to $x=L$ $x = L$ quasi-statically. Two expansion protocols are analytically solvable (Li et al., 2012):
- (A) Isothermal expansion: After re-thermalizing, the work output per cycle is
$W_{\rm exp} = k_B T \ln\frac{Z(L)}{Z(L/2)},$

and the total heat absorbed is

$Q_{\rm tot} = -\sum_k P_{2k-1}(L)[E_{2k}(L)-E_{2k-1}(L)] + k_BT\ln\frac{Z(L)}{Z(L/2)}.$

(B) Adiabatic expansion plus thermalization: The work and heat are distributed differently, but the same net relations hold.

Removal: The barrier is removed; in the ideal quantum case, this can be performed with no additional work or heat.

2. Information-Theoretic Structure: Classical vs Quantum Information

A rigorous analysis exposes the distinct roles of different information contributions in QSZE thermodynamics (Li et al., 2012):

Classical information $S_c = k_B\ln2$ arises from “which-side” knowledge after measurement. It plays a feedback-control role (determining direction of the subsequent expansion), but does not directly contribute to the net heat absorbed or work produced.
Quantum information is quantified as the difference in level-occupancy entropy before and after the insertion/measurement,

$\Delta S_{\rm quantum} = S_0 - h(p),$

where $h(p)$ is the entropy associated with the post-insertion block-diagonal distribution. Only this quantum information determines the cycle's net heat/work:

$Q_{\rm tot} = T\,\Delta S_{\rm quantum},$

thus fully governing the thermodynamic gain.

In the macroscopic limits ( $L\to\infty$ or $T\to\infty$ ), quantum corrections vanish: $W_1\to0$ , $Z(L)/Z(L/2)\to2$ , $\Delta S_{\rm quantum}\to k_B\ln2$ , fully recovering the classical Szilard result $W_{\rm tot}=k_BT\ln2$ .

3. Generalizations: Statistics, Interactions, and Potentials

3.1. Indistinguishable Particles and Spin

The extension to many-body scenarios introduces Bose or Fermi symmetry (Kim et al., 2010, Zhuang et al., 2015, Cai et al., 2011). For $N$ noninteracting particles, the net extractable work is

$W_{\rm tot} = -k_BT\sum_{m=0}^N f_m\,\ln\frac{f_m}{f_m^*},$

where $f_m$ is the probability for $m$ particles on one side after insertion, and $f^*_m$ is the equilibrium probability after expansion. For bosons, at $T\to0$ , $W_{\rm tot}\to k_BT\ln(N+1)$ . For fermions, Pauli exclusion sharply restricts extractable work: $W_{\rm tot}=0$ for even $N$ , $W_{\rm tot}=k_BT\ln2$ for odd $N$ (Thilagam, 2013, Zhuang et al., 2015).

3.2. Interacting Bosons and Quantum Supremacy

Attractive $N$ -boson engines can surpass the classical one-bit $k_BT\ln2$ bound (Bengtsson et al., 2017). At low $T$ , quantum correlations enhance the probability of clustered occupations, yielding

$W/W_1 > 1,$

and a peak in work output at an intermediate $T$ depending on $N$ and interaction strength $g$ .

3.3. Conventional and Exotic Potentials

QSZEs have been solved for particles in harmonic traps (Davies et al., 2020), fractional power-law potentials, Morse potentials (Islam, 2023), and under generalized uncertainty principles (Chen et al., 2016). In all cases, the core principle remains: work extraction and efficiency depend on quantum partition-function ratios, and the conversion of information into thermodynamic resources is sharply modulated by the spectral structure.

4. Thermodynamic Optimality, Irreversibility, and Landauer Principle

The optimal cycle is achieved by effective reversibility: matching the generalized force on the barrier in forward and (averaged) backward protocols (Jeon et al., 2014). However, fundamental irreversibility persists due to quantum measurement's inherently nonunitary character; this manifests as the necessity of entropy production when localizing superposed quantum states—encoded in Landauer’s principle, which dictates a minimal dissipation of $k_BT\ln2$ per bit of erased information for demon-based or demonless engines (Aydin et al., 2019, Ashrafi et al., 2020). Even when explicit Maxwell's demons are absent, the projection (localization) step entails a thermodynamic cost at least as large as the extracted work, preserving the second law.

5. Certification of Quantumness and Experimental Realizations

QSZEs permit device-independent certification of quantum effects via work/entropy inequalities inaccessible to local hidden state models (Beyer et al., 2019). If the average work extracted violates a steering-type bound, a quantum Maxwell demon has been realized.

Experimental demonstrations have spanned NMR qubit systems (Peterson et al., 2020), quantum dots (Barker et al., 11 Nov 2025), and superconducting circuits (Tang et al., 2024), all achieving high-fidelity conversion of theoretically possible information-to-work conversion bounds, and validating generalized fluctuation theorems and thermodynamic uncertainty relations.

6. Outlook and Physical Significance

Quantum Szilard engines crystallize the thermodynamics of information at the quantum scale. By unifying partition function analytics, quantum measurement theory, and the physics of entropy and information erasure, QSZEs have become platforms for probing the limits of the second law, benchmarking quantum statistical effects, and inspiring the engineering of information-powered nanoscopic devices. In future, they will continue to clarify the operational meaning of quantum information and the ultimate limits of energy conversion in quantum technologies (Li et al., 2012, Barker et al., 11 Nov 2025).