Quantum Thermodynamics: An introduction to the thermodynamics of quantum information

Prologue

• Thermodynamics is presented as a phenomenological theory describing average heat and work through a universal framework for the typical behavior of physical systems.Unlike microscopic theories, it does not seek detailed predictions for individual physical systems.Source: Prologue · p. 9
• Quantum Thermodynamics applies thermodynamic principles to quantum technologies by identifying new canonical variables and translating universal thermodynamic concepts into practical optimization statements.The motivating goal is to minimize wasted resources such as work and information in complicated quantum technologies.Source: Prologue · pp. 10–11
• The book provides a concise introduction to Quantum Thermodynamics and its application to quantum systems that process information.It arose from lecture notes for graduate statistical-physics classes, workshops, and summer schools, filling a perceived gap in comprehensive teaching material.Source: Prologue · p. 11

The principles of modern thermodynamics

• Thermodynamics describes the average behavior of heat and work using five axioms known as the laws of thermodynamics.As an axiomatic theory, it remains valid when its basic assumptions are fulfilled.Source: The principles of modern thermodynamics · p. 13
• The theory lacks microscopic information, cannot characterize non-equilibrium states beyond quasistatic processes, and is ill-equipped for direct application to quantum systems.These limitations arise from thermodynamics’ phenomenological, equilibrium, and classical frameworks.Source: The principles of modern thermodynamics · p. 13
• The chapter develops thermodynamics, stochastic thermodynamics, quantum-information characterizations of equilibrium, and quantum thermodynamics in successive sections.These topics are covered in Sections 1.1 through 1.4.Source: The principles of modern thermodynamics · p. 13

1.1 A phenomenological theory of heat and work

• Work is controllable energy transfer, heat is energy exchanged with a bath, and internal energy is a state function independent of the transformation path.Work and heat are non-exact differentials, whereas internal-energy changes are exact.Source: §1.1 · p. 13, §1.1.1 · pp. 15–16

1.1.1 The five laws of thermodynamics

• Thermodynamics axiomatizes transformations between equilibrium states, characterized by experimentally accessible parameters linked through an equation of state.For an ideal gas, the equation of state is PV = NkBT; equilibrium states form a smooth thermodynamic manifold.Source: §1.1.1 · pp. 13–14
• Quasistatic processes remain on the thermodynamic manifold and are reversible, whereas finite-rate processes leave it through nonequilibrium states whose irreversibility must be quantified.This need to quantify irreversibility motivates stochastic thermodynamics.Source: §1.1.1 · pp. 14–15

1.1.2 Finite-time thermodynamics and endoreversibility

• Standard Carnot, Otto, and Diesel cycles have zero power because they consist solely of quasistatic transformations, motivating finite-time and endoreversible thermodynamics.Endoreversibility treats the working medium as locally reversible while heat-transfer gradients make the overall cycle irreversible.Source: §1.1.2 · pp. 18–19
• The endoreversible Carnot cycle has a universal efficiency at maximal power that depends only on the hot and cold reservoir temperatures [8].The working medium remains locally equilibrated while remaining globally irreversible relative to the environment.Source: §1.1.2 · p. 20

1.2 The advent of Stochastic Thermodynamics

• Fluctuation theorems generalize the second law to far-from-equilibrium systems: negative entropy-production events occur, but are exponentially unlikely, while average entropy production retains the conventional inequality.The fluctuation theorem relates probabilities of opposite entropy-production values and emerged from simulations of small, fluctuation-dominated systems; its first rigorous proof was given by Gallavotti and Cohen in 1995.Source: §1.2 · pp. 20–21

1.2.1 Microscopic dynamics

• Microscopic dynamics can be modeled by Langevin equations for stochastic trajectories or equivalent Klein-Kramers equations for phase-space probability densities, with the latter enabling direct calculation of entropy production.The Langevin description uses damping, a confining potential, and Gaussian white noise; its equilibrium limit yields the Fluctuation-Dissipation Theorem, while the Klein-Kramers stationary state is Boltzmann-Gibbs.Source: §1.2.1 · pp. 21–22

1.2.2 Stochastic energetics

• Stochastic energetics assigns internal-energy, heat, and work increments to individual trajectories, producing a microscopic form of the first law under overdamped dynamics driven by an external control parameter λ.The overdamped assumption treats momentum as equilibrated on faster timescales, and the work increment is determined by the potential’s derivative with respect to λ.Source: §1.2.2 · p. 23

1.2.3 Jarzynski equality and Crooks theorem

• The Jarzynski equality extends the second law across broad nonequilibrium conditions, whereas Crooks’s theorem gives a detailed forward–reverse work relation but requires Markovian dynamics.Crooks’s trajectory construction identifies parameter-driven energy changes as work and environment-induced state-transition changes as heat; the Jarzynski equality also holds for non-Markovian dynamics.Source: §1.2.3 · pp. 25–26

1.3 Foundations of statistical physics from quantum entanglement

• Standard derivations of canonical thermal equilibrium rely on concepts that are not well-phrased for quantum systems, while infinite-copy ensembles are problematic in a fully quantum theory.Statistical physics originally translated between classical microstates and thermodynamic macrostates using ensembles.Source: §1.3 · p. 27
• Microcanonical and canonical equilibrium representations can be derived from quantum entanglement symmetry through entanglement-assisted invariance, or envariance.This approach provides a fully quantum treatment of the foundations of statistical mechanics.Source: §1.3 · p. 27

1.3.1 Entanglement assisted invariance

• Envariance means that an environment-only unitary can undo a system-only unitary on a maximally entangled composite state, restoring the global state without acting on the system.For a system unitary US, an environment unitary UE exists such that UE|ψSE⟩=|ψSE⟩.Source: §1.3.1 · p. 27
• For an entangled pair of two-level systems, an environment swap restores the pre-swap global state and leaves the swapped local probabilities equal, connecting quantum states and probabilities and leading to Born’s rule.The argument uses a system spin swap counteracted solely by a corresponding environment swap.Source: §1.3.1 · pp. 27–28
• Experiments in quantum optics and on IBM’s Q Experience show that envariance is physically realizable, supporting its use in deriving quantum equilibrium states.These observations establish envariance as more than a theoretical concept.Source: §1.3.1 · p. 28

1.3.2 Microcanonical state from envariance

• The microcanonical equilibrium is defined as a quantum state maximally envariant under all system unitaries, which occurs exactly when all Schmidt coefficients have equal absolute value [60].Thus, the Schmidt decomposition is even: |ak|=|al| for all k and l.Source: §1.3.2 · p. 28
• The quantum microcanonical equilibrium is an equivalence class of states that are both envariant under all system unitaries and fully energetically degenerate with respect to the Hamiltonian.The degeneracy is defined within the energy-dependent Hilbert-space dimension Zmic, the microcanonical partition function.Source: §1.3.2 · p. 28

1.3. FOUNDATIONS OF STATISTICAL PHYSICS FROM QUANTUM ENTANGLEMENT 21

• Envariance replaces ambiguous notions such as probability, ergodicity, ensembles, randomness, and indifference in the reformulation of statistical mechanics.The approach uses quantum symmetry induced by entanglement rather than relying on these incompletely defined concepts.Source: §1.3 · p. 29
• Microcanonical equilibrium is defined as a fully energetically degenerate quantum state envariant under all unitaries.The treatment also extends this fully quantum-mechanical foundation to canonical equilibrium.Source: §1.3 · p. 29

1.3.3 Canonical state from quantum envariance

• The canonical derivation counts bath degeneracies for subsystem energy eigenstates, using ultraweak coupling and an energy spectrum at least as dense as the subsystem’s.Envariance permits choosing subsystem states as energy eigenstates, while the interaction is neglected in the total energy and composite equilibrium state.Source: §1.3.3 · pp. 29–30
• The qubit example gives bath degeneracy by counting distributions of n ground-state and N−n excited-state qubits, assuming non-interacting bath qubits.The construction explicitly depends on subsystem energy eigenstates and the absence of subsystem–bath interactions.Source: §1.3.3 · p. 30
• The generalized counting formula is a quantum-envariant version of Boltzmann’s microstate counting, and variational maximization yields the Boltzmann-Gibbs formula.The derivation uses Stirling’s approximation for N≫1 and interprets temperature through the Lagrange multiplier β.Source: §1.3.3 · pp. 30–31
• A unique β establishes an envariance-based zeroth law: systems equilibrated with a large bath equilibrate with each other at the same temperature.The discussion is exact apart from Stirling’s approximation, but deriving the canonical formula additionally requires thermodynamic conditions on reservoir-state occupation.Source: §1.3.3 · p. 31

1.4.1 Quantum work and quantum heat

• For Gibbs equilibrium states, heat is internal-energy change associated with entropy, whereas work is the change caused by varying the system Hamiltonian.The Gibbs entropy satisfies S = β(E − F), and internal-energy changes separate into useful free-energy work and the entropic contribution T dS.Source: §1.4.1 · p. 32
• Non-Gibbsian equilibrium requires separating entropic excess heat from correlation-maintenance costs, while the thermodynamically consistent quantum Carnot efficiency remains identical to the classical Carnot efficiency.Energetic back action maintains coherence and system–environment correlations; the excess heat alone carries the entropic cost.Source: §1.4.1 · pp. 33–34

1.4.2 Quantum entropy production

• Equation (1.98) gives the exact microscopic mean entropy production for a driven open quantum system weakly coupled to one heat reservoir, even arbitrarily far from equilibrium.The expression applies along nonequilibrium paths generated by varying the external control parameter.Source: §1.4.2 · p. 35

1.4.3 Two-time energy measurement approach

• The two-time energy-measurement approach makes work stochastic and satisfies a quantum Jarzynski equality, but projective measurements introduce informational work costs and make work path dependent rather than an observable.A measurement-free alternative avoids the observer’s thermodynamic payment, while the modified Jarzynski equality and maximum-work theorem account for measurement-related information.Source: §1.4.3 · pp. 35–37

1.4.4 Quantum fluctuation theorem for arbitrary observables

• Because projective energy measurements are often infeasible, the framework is generalized toward fluctuation theorems based on measurements of arbitrary accessible observables.The generalized protocol measures an observable at the beginning and end of the process, such as spatial density or magnetization.Source: §1.4.4 · p. 39

m. As before, Πi

• The measurement protocol applies a CPTP evolution between measurements of initial and final observables, generating a joint distribution of outcomes and the random change ∆ω.Noncommuting initial states and observables mean that measurement can alter the system’s statistics.Source: m. As before, Πi · pp. 39–40
• The quantum efficacy ε quantifies information lost by not measuring the environment and is explicitly determined by the map E.The resulting relation is therefore not an integral fluctuation theorem in the strict sense unless ε is independent of measurement details.Source: m. As before, Πi · p. 40
• The quantum Jarzynski relation becomes a fluctuation theorem for Gibbs initial states, energy measurements, and unital dynamics, while remaining valid for purely decohering or dephasing models.More generally, its quantum efficacy depends on the observables, initial state, and CPTP map, so it is not protocol-independent.Source: m. As before, Πi · p. 40

1.4.5 Quantum entropy production in phase space

• A quantum Liouvillian need not generally exist for all quantum systems, and for a thermally open harmonic oscillator its existence and explicit form depend on the environment’s initial preparation.For an open quantum system in equilibrium, the stationary Wigner function is generally not the Wigner representation of the Gibbs state; a harmonic oscillator example is known and solvable analytically.Source: §1.4.5 · p. 41
• For any open or closed quantum system described by the phase-space master equation, the entropy production defined through quantum trajectories satisfies an integral fluctuation theorem.The trajectory is a mathematical construct, with physical quantities obtained by ensemble averages, and the proof uses an exponentially weighted marginal and the stationary Wigner function.Source: §1.4.5 · pp. 41–42

1.5 Checklist for “The principles of modern thermodynamics”

• The checklist characterizes thermodynamics as a phenomenological theory of average heat and work, with reversible processes represented as paths on a thermodynamic manifold.It also distinguishes local equilibrium from global equilibrium and permits trajectory-level definitions of heat, work, and entropy in classical systems.Source: §1.5 · p. 43
• The checklist states that fluctuation theorems express exponentially unlikely violations of the second law, while quantum work is not an ordinary observable and admits multiple justified definitions.It also notes that statistical mechanics can be formulated from a quantum framework using symmetries of entanglement.Source: §1.5 · p. 43

A phenomenological theory of heat and work 1.1

• The exercises apply the framework to a particle in an infinite square well and to a quantum Otto engine using a harmonically trapped particle as the working medium.They ask for equations of state, thermodynamic manifolds, particle-statistics comparisons, and Otto-cycle efficiency under Gibbs-state assumptions.Source: A phenomenological theory of heat and… · p. 43

The advent of stochastic thermodynamics 1.2

• Stochastic-thermodynamics problems analyze work distributions for classical harmonic oscillators under driven dynamics and test fluctuation relations.The supplied exercises use Maxwell–Boltzmann initial states and ask for verification of Jarzynski or Crooks relations.Source: The advent of stochastic… · p. 43
• [3] A one-dimensional classical harmonic oscillator follows the classical Liouville equation while its angular frequency varies, requiring the work density and verification of the Jarzynski equality.The oscillator is initially prepared in a Maxwell–Boltzmann distribution.Source: The advent of stochastic… · p. 43
• [4] A one-dimensional classical harmonic oscillator coupled to a thermal bath follows the classical Klein–Kramers equation while being dragged along the x-axis, requiring the work density and verification of the Crooks fluctuation theorem.The oscillator is initially prepared in a Maxwell–Boltzmann distribution.Source: The advent of stochastic… · p. 43

• An exercise asks readers to illustrate envariance for a universe consisting of two harmonic oscillators [5].Source: Foundations of statistical physics… · p. 44
• Another exercise concerns parity-preserving unitary maps.Source: Foundations of statistical physics… · p. 44
• Readers are asked to repeat the arguments leading to Eq. (1.78) while including the next two terms of the Stirling expansion.The supplied expression includes 2 ln(2πn) + 1/(12n).Source: Foundations of statistical physics… · p. 44
• A final question asks how temperature would be identified in this case.Source: Foundations of statistical physics… · p. 44

Quantum work

• [7] asks for the work probability density during an infinitely slow angular-frequency variation of a thermally isolated quantum harmonic oscillator initially in a Gibbs state.The oscillator is one-dimensional and harmonic.Source: Work, quantum heat, and quantum… · p. 44

Quantum entropy production

• [8] asks for the entropy-production probability density when a bath-coupled one-dimensional quantum harmonic oscillator is dragged along the x-axis.Its dynamics follow the quantum Klein-Kramers equation (1.131).Source: Work, quantum heat, and quantum… · p. 44
• The bath-coupled oscillator exercise also asks students to verify the quantum fluctuation theorem (1.139).This verification accompanies the entropy-production calculation.Source: Work, quantum heat, and quantum… · p. 44

2.1 Quantum thermometry

• Quantum thermometry uses quantum estimation theory because temperature is generally neither a classical nor quantum observable, with QFI defining the ultimate measurement sensitivity.A probe equilibrates with the system of interest, is decoupled, and is measured to estimate temperature.Source: §2.1 · pp. 49–50, Figure 2.1 · p. 50

2.1.1 Thermometry for Harmonic Spectra

• For harmonic spectra, smaller energy gaps improve low-temperature precision, while each thermometer’s QFI peaks at one characteristic temperature set by its gap.Two-level and harmonic-oscillator thermometers show qualitatively identical behavior, indicating that the gap matters more than dimensionality.Source: §2.1.1 · p. 51, Figure 2.2 · p. 52
• Harmonic thermometers perform identically at low temperatures, gain only modestly from a three-level system, and obtain no further optimal-precision advantage from higher dimensions.Larger gaps are more useful at higher temperatures, but these systems remain optimized for estimating a single temperature.Source: §2.1.1 · p. 52

2.1.2 Optimal Thermometers

• Optimal thermometers maximize equilibrium energy variance or heat capacity through an effective two-level structure with a degenerate excited manifold.Higher dimensionality improves precision but narrows the temperature range, whereas anharmonicity can broaden the range of temperatures probed precisely.Source: §2.1.2 · pp. 52–53

2.2 Quantum heat engines – engines with atomic working fluids

• The quantum Otto cycle is studied with genuinely quantum working substances, especially a two-level system and a quantum harmonic oscillator, including an experimental realization of the latter.The cycle is introduced because it is both practically relevant and a central cyclic process in classical and quantum thermodynamics.Source: §2.2 · p. 53

2.2.1 The Otto Cycle: Classical to quantum formulation

• Quantum Otto strokes separate work and heat: adiabatic changes vary energy levels at fixed populations, whereas isochoric changes vary populations at fixed energy levels.Quantum adiabatic strokes must be quasistatic to avoid excitations, unlike rapid classical adiabatic processes, and Otto thermalization is irreversible unlike ideal Carnot isotherms.Source: §2.2.1 · pp. 54–55

2.2.2 A two-level Otto cycle

• A two-level quantum Otto cycle performs work during isentropic compression and expansion and exchanges heat during isochoric heating and cooling, while remaining in bath equilibrium only at A and C.Because the working substance must rethermalize with the baths, the cycle is irreversible; its efficiency is governed by κ ≡∆i/∆f and remains below Carnot efficiency.Source: §2.2.2 · pp. 57–58, Figure 2.4 · p. 57

2.2.3 Endoreversible Otto cycle

• Quantum engines emphasize efficiency at maximum power rather than only maximum efficiency, motivating the endoreversible Otto cycle and its relation to the Curzon-Ahlborn benchmark.The section introduces this focus after noting the close similarities between quantum and classical Otto cycles.Source: §2.2.3 · p. 58
• The classical harmonic oscillator’s endoreversible Otto efficiency at maximal power equals the Curzon–Ahlborn efficiency (1.22).This result follows analytically after eliminating thermodynamic parameters using Fourier heat transfer and harmonic-oscillator entropy and energy expressions.Source: §2.2.3 · p. 61
• In the quantum harmonic engine, high-temperature efficiency again equals Curzon–Ahlborn, whereas deep-quantum efficiency is larger.The quantum power expression no longer factorizes, so the maximizing compression ratio depends on stroke times; the regimes are illustrated in Figure 2.5.Source: §2.2.3 · p. 62, Figure 2.5 · p. 62

2.3 Work extraction from quantum systems

• Useful work from a quantum battery requires a cyclic unitary transformation that lowers its final energy while restoring the Hamiltonian.Passive states cannot yield work under such processes, while active states can; the extractable amount is called ergotropy.Source: §2.3 · pp. 63–64

2.3.1 Work extraction from arrays of quantum batteries

• For qubit battery arrays, passive states remain completely passive, so collective extraction cannot outperform the local classical strategy.The maximal ergotropy of two qubits equals twice the single-qubit value, confirming the classical-strategy bound.Source: §2.3.1 · pp. 64–65
• A direct non-local swap extracts maximal two-qubit ergotropy three times faster than a non-entangling protocol, although intermediate entanglement is unnecessary for maximal work.The faster protocol may generate entanglement, whereas the slower multi-step protocol remains separable; non-entangling extraction can still generate quantum discord.Source: §2.3.1 · p. 65
• For anharmonic batteries with dimension d > 2, passive states need not be completely passive, enabling collective entangling operations to outperform local extraction under suitable conditions.Reaching the completely passive state in this regime necessarily involves sharing classical correlations between batteries.Source: §2.3.1 · p. 65
• Under suitable conditions, global processing of multiple quantum batteries extracts more work than local processing, can reach maximal ergotropy without entanglement, and increases power when entangling operations reduce the number of swaps.For p1 > pth, an additional swap produces classical correlations and can exceed the classical limit; global operations can instead generate strong quantum correlations during faster discharge.Source: §2.3.1 · pp. 66–67

2.3.2 Powerful charging of quantum batteries

• Global entangling charging operations reduce the charging time of an array and yield an n-fold increase in charging power under the same energy constraint.The charging process is equivalent to reversing the work-extraction operation, with the speed advantage arising from a global swap between the collective ground and maximally excited states.Source: §2.3.2 · pp. 67–68

2.4.1 Work, heat, and entropy production for dynamical semigroups

• Quantum dynamical semigroups model memoryless open-system evolution; for a thermal bath, the qubit dynamics has a Gibbs stationary state, and with a time-independent Hamiltonian energy changes are entirely heat exchange.The irreversible entropy production is positive, while non-Markovian dynamics makes its definition significantly more difficult; non-Gibbsian invariant states can arise depending on the Hamiltonian and interactions.Source: §2.4.1 · pp. 68–70

2.4.2 Entropy production as correlation

• For finite-dimensional system-environment dynamics, the system’s irreversible entropy production is identified with correlations established between system and environment when the environment remains in equilibrium.The system entropy change separates into reversible heat-related and irreversible contributions; under an invariant environment, the latter equals the system-environment correlations up to sign.Source: §2.4.2 · pp. 70–71

2.4.3 Quantum Darwinism: Emergence of classical objectivity

• Quantum decoherence selects pointer states through environment-induced superselection: pointer states remain stable, whereas their superpositions lose phase coherence.The selected states depend on the initial system-environment states and their interaction details.Source: §2.4.3 · p. 72
• Quantum Darwinism explains classical objectivity through redundant environmental encoding, so different observers accessing separate environmental fractions infer the same system state.In the qubit model, the system decoheres without changing populations, those populations are imprinted onto environmental qubits, and the mutual-information plateau signals redundancy; Figure 2.6 shows this plateau for N = 16.Source: §2.4.3 · pp. 72–73, Figure 2.6 · p. 74
• At t = π/(4J), the system is fully decohered and every environmental fragment shares the system entropy, producing a mutual-information plateau that signals redundant encoding.The spin-star interaction leaves populations unchanged while suppressing coherences; quantum Darwinism explains classical objectivity through repeated environmental imprints.Source: §2.4.3 · pp. 74–75, §2.5 · p. 76

2.5 Checklist for “Thermodynamics of Quantum Systems”

• Quantum systems with maximal energy variance, and therefore heat capacity, are the optimal thermometers.The checklist notes that measuring low temperatures is not as simple as inserting a thermometer into a quantum system.Source: §2.5 · p. 76
• Reversible Otto engines with quantum working fluids achieve the same efficiency as their classical counterparts.The accompanying problem asks for the efficiency of a quantum Otto engine under quasistatic magnetic-field variation.Source: §2.5 · p. 76, Quantum heat engines – engines with… · p. 76
• Quantum working fluids can exceed the Curzon-Ahlborn efficiency at maximum power for endoreversible Otto cycles.A related problem considers the endoreversible maximum-power efficiency of an Otto engine with a semiclassical ideal-gas working medium.Source: §2.5 · p. 76, §2.6 · p. 77
• Entangling operations extract work from multiple copies of passive but not completely passive quantum batteries, while dynamical entanglement boosts the charging power of battery arrays.The section also frames battery ergotropy relative to completely passive states as an exercise topic.Source: §2.5 · p. 76, Work extraction from quantum systems 2.3 · p. 77

Quantum decoherence and the tale of quantum Darwinism 2.4

• Entropy production can be understood as correlations established between a quantum system and its environment.The decoherence exercises examine pure decoherence, Lindblad form, unital dynamics, and whether an entangled state can still exhibit quantum-Darwinist objectivity.Source: Quantum decoherence and the tale of… · p. 77

Thermodynamics of Quantum Information

• Realistic quantum information processing produces irreversible entropy and irretrievably loses quantum information through environmental decoherence and dissipation.The chapter therefore examines thermodynamic costs, adiabatic quantum-computer performance, defect formation, and error correction.Source: Thermodynamics of Quantum Information · p. 82

3.1.1 Thermodynamics of classical information processing

• Landauer’s principle establishes an absolute thermodynamic minimum for erasing information, while reversible computation can avoid intrinsic costs from logically reversible operations.Logical irreversibility maps multiple inputs to one output; reversible embeddings such as Toffoli gates can avoid that mapping.Source: §3.1.1 · p. 83
• Erasing a bit decreases its information entropy, so the environment must undergo an equal or greater entropy increase.This contrasts with faithful copying of a definite bit, which can be performed reversibly without changing entropy.Source: §3.1.1 · p. 83
• Landauer’s erasure argument links information entropy to thermodynamic entropy and bounds the heat deposited in a reservoir during erasure.For a two-state bit, the initial Shannon entropy is ln2 and the erased state has zero entropy.Source: §3.1.1 · pp. 83–84
• kBT ln2 of work is required to erase a bit in the double-well protocol.The protocol slowly removes the barrier, biases the wells, and restores the barrier; for a definite initial state, the work becomes environmental heat.Source: §3.1.1 · p. 84

Maxwell’s Demon and Szilard’s Engine.

• In the Szilard engine, extracting kBT ln2 of work is reconciled with the second law because erasing the Demon’s memory dissipates kBT ln2 of heat, whereas measurement itself has no intrinsic cost.The Demon sorts particles by measurement, but completing the cycle requires returning its memory to a standard state.Source: Maxwell’s Demon and Szilard’s Engine · pp. 85–86

3.1.2 A quantum sharpening of Landauer’s bound

• Under quantum-system, initially uncorrelated, thermal-environment, and globally unitary assumptions, the non-equilibrium quantum Landauer principle follows from entropy production and system–environment mutual information.Initial correlations can reduce system entropy without a corresponding environmental increase, while global unitarity accounts for all resources used.Source: §3.1.2 · pp. 86–87
• Equality in the rigorous Landauer bound holds only for trivial processes, and the treatment relates system entropy change directly to heat transferred to the reservoir.This quantum formulation is valid for non-equilibrium settings and distinguishes entropy-change reasoning from entropy-production derivations.Source: §3.1.2 · p. 87

3.1.3 New Landauer bounds for non-equilibrium quantum systems

• Full counting statistics and a two-time environmental-energy measurement protocol provide an alternative, thermodynamic family of lower bounds on mean dissipated heat.The approach assumes an initially uncorrelated system and thermal environment, and derives bounds from the convexity of the cumulant generating function.Source: §3.1.3 · p. 88, §3.2 · p. 89

3.2.1 Fluctuation theorem for quantum annealers

• Quantum fluctuation theorems yield experimentally accessible diagnostics for annealers by relating the final energy-outcome distribution to whether evolution remains ideal, unitary, and adiabatic.The framework uses a one-dimensional transverse-field quantum Ising model and applies for arbitrary annealing durations when computation starts and finishes in ground states.Source: §3.2.1 · p. 90, §3.2 · p. 91

3.2. PERFORMANCE DIAGNOSTICS OF QUANTUM ANNEALERS 83

• For useful but nonideal annealing, every dependence of the outcome distribution on annealing time indicates dissipation or decoherence through environmental interactions.Ideal adiabatic evolution maps the initial paramagnetic ground state to a defect-free final ground state, whereas hardware imperfections can make the dynamics nonunitary or nonadiabatic.Source: §3.2.1 · p. 90, §3.2 · p. 91

3.2.2 Experimental test on the D-Wave machine

• On D-Wave 2X and 2000Q machines, the measured outcome distributions generally deviated from the theoretical prediction, depended on annealing time, and indicated substantial computational errors.Departures from the ideal distribution correspond to kinks or topological defects, allowing the annealer’s miss of the true ground state to be quantified.Source: §3.2.2 · pp. 92–93, Figure 3.3 · p. 92, Figure 3.4 · p. 93

3.3 Kibble-Zurek Scaling of Irreversible Entropy

• The Kibble–Zurek mechanism explains defect formation near continuous phase transitions through critical slowing down, divergent correlation length and time, and universal critical behavior.Its scaling uses a distance-from-criticality parameter ε with spatial and dynamical critical exponents ν and z; ε represents relative temperature or external field depending on the transition.Source: §3.3 · pp. 93–94, §3.3.1 · p. 94

3.3.1 Fundamentals of the Kibble-Zurek mechanism

• Finite-rate driving freezes near criticality, and Kibble–Zurek scaling expresses defect production through equilibrium critical exponents.The dynamics is adiabatic far from the critical point but enters an impulse regime near it; defect density or correlation length can quantify the resulting scaling.Source: §3.3.1 · p. 94

3.3.2 Example: the Landau-Zener model

• The Landau–Zener model provides an illustrative test of Kibble–Zurek theory, with defect density determined from finite-rate excitations and almost perfect numerical agreement.The avoided crossing connects adiabatic ground-state evolution to excitation-induced precession, with fidelity quantifying the defect density.Source: §3.3.2 · pp. 95–96

3.3.3 Kibble-Zurek mechanism and entropy production

• For constant-rate critical driving, excess work universally follows KZ scaling and depends on the critical exponent of the varied control parameter.Excess work arises only in the impulse regime, while the typical domain size is independent of which intensive control parameter is varied; for the quantum Ising chain, ⟨Wex⟩∼τQ^-1 [32].Source: §3.3.3 · p. 98

3.4 Error correction in adiabatic quantum computers

• Quantum error correction faces significant challenges from measurement back action, continuous quantum noise, and the inability to copy quantum states.The Shor code protects a single qubit against any arbitrary error, and its utility has been demonstrated in several experiments.Source: §3.4 · p. 99

3.4.1 Quantum error correction in quantum annealers

• Quantum annealing correction combines encoding with an energy penalty to correct environmental noise, but it cannot eliminate finite-time excitations from the ground-state manifold.The scheme was successfully tested for n = 3 on a second-generation D-Wave machine.Source: §3.4.1 · p. 100

3.4.2 Adiabatic quantum computing – A case for shortcuts to adiabaticity

• Transitionless quantum driving constructs an auxiliary Hamiltonian that suppresses nonadiabatic transitions at every instant, making it suited to bypass fundamentally non-correctable annealing errors.Unlike simply slowing the process, it preserves evolution along the adiabatic manifold during finite-time driving.Source: §3.4.2 · p. 101

3.4.3 Counterdiabatic Hamiltonian for scale-invariant driving

• A major limitation of transitionless driving is that closed-form counterdiabatic fields are rarely available because they depend on the full spectral decomposition of H0(t).The difficulty is reduced for a reasonably broad class of systems.Source: §3.4.3 · p. 102
• Transitionless driving permits arbitrary ramps and, in principle, arbitrarily short driving times without imposing constraints on the time dependence.For the LMG model, the method yields exact correction terms for perfect finite-time adiabatic dynamics as N →∞.Source: §3.4.3 · pp. 102, 105
• For scale-invariant Hamiltonians, the counterdiabatic term has a closed form proportional to the dilation generator {q, p}, independent of explicit energy eigenfunctions.This applies to non-stationary superpositions and mixed states, although the resulting Hamiltonian is non-local and difficult to realize experimentally.Source: §3.4.3 · p. 103
• Transitionless driving is limited computationally because non-scale-invariant protocols require instantaneous eigenstates, while resulting auxiliary Hamiltonians can be non-local, singular at criticality, and inaccurate for finite N.The LMG correction is not well-defined at the critical point and does not exactly recover adiabatic dynamics when applied verbatim to finite systems.Source: §3.4.3 · pp. 105–106

3.5 Checklist for “Thermodynamics of Quantum Information”

• Information processing has thermodynamic energy costs, while coherence and entanglement provide additional quantum thermodynamic resources.The checklist identifies information writing and erasure as costly processes and purely quantum effects as usable resources.Source: §3.5 · p. 107
• Quantum stochastic thermodynamics assesses quantum-computer performance, and quantum entropy production quantifies computational errors, including Kibble-Zurek scaling and phenomenological error prediction.The checklist also identifies quantum error correction as essential and shortcuts to adiabaticity as suppressing fundamentally non-correctable errors in quantum annealers.Source: §3.5 · p. 107

• [1] The double-well problem asks whether resetting a Maxwell–Boltzmann-distributed system into the right well with error probability below δ satisfies Landauer’s principle.The final state is modeled as a narrow Gaussian centered in the right well, with left-well probability smaller than δ.Source: Quantum thermodynamics of information… · p. 107
• [2] The two-qubit problem considers an even mixture of Bell states and asks for the heat dissipated when the stored quantum information is completely erased.Source: Quantum thermodynamics of information… · p. 107
• The erasure analysis separately asks how much dissipated heat comes from classical information versus destroying quantum correlations.Source: Quantum thermodynamics of information… · p. 107

Performance diagnostics of quantum annealers 3.2

• [3] Formulate the dynamics of a two-level system weakly coupled to thermal noise using a Lindblad master equation.Source: Performance diagnostics of quantum… · p. 108
• Compute the quantum efficacy ε for Ai = σx, Af = σx, and an initially thermal state ρ0 = exp(−βH)/Z.The Hamiltonian is H = −ℏωσz/2.Source: Performance diagnostics of quantum… · p. 108
• [4] Analyze the strong-coupling master equation with a σz system–bath operator for H = −ℏωσz/2.Source: Performance diagnostics of quantum… · p. 108
• Show that the dynamics generated by the strong-coupling master equation is unital.Source: Performance diagnostics of quantum… · p. 108

Kibble-Zurek Scaling of Irreversible Entropy 3.3

• [5] For a Landau–Zener Hamiltonian, compute the quantum work distribution for evolution from ti = −t̂ to tf = t̂, starting in |↑⟩.The exercise asks to show that the excess work exhibits Kibble–Zurek scaling.Source: Kibble-Zurek Scaling of Irreversible… · p. 108
• [6] Use mean-field Landau free energy to identify the critical point and critical exponents, then predict irreversible entropy production under finite-rate driving through criticality.Source: Kibble-Zurek Scaling of Irreversible… · p. 108

Error correction in adiabatic quantum computers 3.4

• The section asks readers to consider two coupled qubits described by an Ising Hamiltonian in a transverse field [7].Source: Error correction in adiabatic quantum… · p. 108
• It asks readers to compute the auxiliary Hamiltonian H1(t) in Eq. (3.58) for transitionless quantum driving.Source: Error correction in adiabatic quantum… · p. 108
• For n = 2 and ⟨N⟩ = 3, it asks readers to derive H1(t) for quantum annealing correction and assess whether Htot(t) could be implemented on a D-Wave annealer.Source: Error correction in adiabatic quantum… · p. 108

Epilogue

• The epilogue surveys thermodynamics from axiomatic origins through stochastic thermodynamics to quantum formulations, emphasizing quantum work, heat, and the importance of understanding thermodynamics for advancing technologies.Source: Epilogue · p. 114
• It identifies resource theories as an uncovered topic, explaining that entanglement and quantum correlations are resources and that thermodynamic resource theories clarify free states, operations, and information costs.Source: Epilogue · p. 114
• The authors argue that thermodynamic principles are crucial for practical, energy-efficient quantum devices, while substantial work remains before quantum technologies fulfill their promise.Source: Epilogue · p. 114