Thermodynamics

Status: Stable core; active frontier at the gravity/information/nonequilibrium interfaces. Last updated: 2026-06-08

Thermodynamics is the macroscopic theory of energy, heat, work, and entropy for systems at or near equilibrium, together with its statistical-mechanical foundations and its modern extensions into fluctuation theorems, information thermodynamics, and gravitational (black-hole) thermodynamics. It is the discipline that converts a small set of phenomenological laws into exact, substance-independent constraints on what physical processes are possible.

---

Scope

This page covers the macroscopic theory of energy, heat, work, and entropy for systems at or near equilibrium: the zeroth through third laws; state functions and the Legendre-transform structure relating internal energy to the free energies; the Carnot bound and reversibility/irreversibility; the second law as the thermodynamic arrow of time and its statistical reinterpretation; the far-from-equilibrium fluctuation theorems (Jarzynski, Crooks) that sharpen the second-law inequality into an equality; Landauer's principle tying logical irreversibility to a minimum dissipation; the Bekenstein–Hawking entropy and Hawking temperature with the generalized second law; and the status of negative and bounded-spectrum temperatures.

It does not derive the full microscopic ensemble theory — that is the job of statistical mechanics — except where the micro–macro link is load-bearing. Kinetic theory and transport (the Boltzmann equation, Green–Kubo relations) are treated only at the interface. Detailed materials-science and chemical-engineering applications are out of scope.

---

Core formalism

1. State space and the fundamental relation

ESTABLISHED A simple one-component system is described by an extensive triple $(U,V,N)$ — internal energy, volume, particle number. Callen's postulatory formulation asserts the existence of an entropy $S = S(U,V,N)$, the fundamental relation, which is (i) single-valued, continuous, and differentiable; (ii) first-order homogeneous (extensive), $S(\lambda U,\lambda V,\lambda N)=\lambda S(U,V,N)$; (iii) monotonically increasing in $U$; and (iv) maximized over unconstrained internal variables at equilibrium. Everything thermodynamic about the system is encoded in this one function. Equivalently, one works in the energy representation $U=U(S,V,N)$, which is convex in its arguments.

The differential form is the Gibbs relation, $dU = T\,dS - p\,dV + \mu\,dN,$ which simultaneously defines the conjugate intensive variables $T=\left(\frac{\partial U}{\partial S}\right)_{V,N},\qquad p=-\left(\frac{\partial U}{\partial V}\right)_{S,N},\qquad \mu=\left(\frac{\partial U}{\partial N}\right)_{S,V}.$ In the entropy representation this reads $dS = \tfrac{1}{T}\,dU + \tfrac{p}{T}\,dV - \tfrac{\mu}{T}\,dN$, and the temperature appears in its most foundational guise: $\frac{1}{T}=\left(\frac{\partial S}{\partial U}\right)_{V,N}.$ ESTABLISHED Because $S$ is first-order homogeneous, Euler's theorem gives the Euler relation $U = TS - pV + \mu N$; differentiating it against the Gibbs relation yields the Gibbs–Duhem relation $S\,dT - V\,dp + N\,d\mu = 0,$ which expresses that the intensive variables $(T,p,\mu)$ are not independent — fixing two constrains the third.

2. The laws as axioms

• Zeroth law ESTABLISHED. Thermal equilibrium is an equivalence relation; its transitivity ("if $A\sim C$ and $B\sim C$ then $A\sim B$") guarantees a consistent empirical temperature labeling the equilibrium classes. This is logically prior to the definition of $T$ as a derivative.
• First law ESTABLISHED. $dU = \delta Q + \delta W$ (energy conservation including heat). $U$ is an exact differential (state function); $\delta Q$ and $\delta W$ are inexact (path-dependent). For quasi-static processes $\delta Q_{\rm rev}=T\,dS$ and $\delta W_{\rm rev}=-p\,dV$.
• Second law ESTABLISHED, with equivalent statements. Clausius: no cyclic process whose sole effect is heat transfer from a colder to a hotter body. Kelvin–Planck: no cyclic process whose sole effect is the complete conversion of heat into work. Entropy form (Clausius–Carathéodory): there exists a state function $S$ with $dS \ge \delta Q/T$ (equality iff reversible), and for an isolated system $\Delta S \ge 0$. For a cycle, the Clausius inequality $\oint \delta Q/T \le 0$.
• Third law ESTABLISHED, with caveats. Nernst's heat theorem: entropy changes of isothermal processes vanish as $T\to 0$. The stronger Planck statement fixes $S\to S_0$ (conventionally $0$) as $T\to 0$ for a perfect crystal. The unattainability form: no finite sequence of operations reaches $T=0$. The caveats (residual entropy, glasses) are domain mismatches discussed below.

3. Thermodynamic potentials and the Legendre structure

ESTABLISHED The free energies are Legendre transforms of $U(S,V,N)$ that trade an extensive variable for its conjugate intensive partner, making the natural control variables the ones an experimenter actually fixes: $F = U - TS \ \ (T,V,N),\qquad H = U + pV \ \ (S,p,N),$ $G = U - TS + pV \ \ (T,p,N),\qquad \Omega = U - TS - \mu N \ \ (T,V,\mu).$ Their differentials, $dF = -S\,dT - p\,dV + \mu\,dN,\quad dG = -S\,dT + V\,dp + \mu\,dN,\quad d\Omega = -S\,dT - p\,dV - N\,d\mu,$ yield the Maxwell relations by equality of mixed second partials — e.g. from $F$, $\left(\partial S/\partial V\right)_T = \left(\partial p/\partial T\right)_V$; from $G$, $\left(\partial S/\partial p\right)_T = -\left(\partial V/\partial T\right)_p$. These exactness constraints relate hard-to-measure entropy derivatives to mechanical equation-of-state data.

ESTABLISHED Extremum principles follow from $\Delta S_{\rm univ}\ge 0$ applied to system-plus-reservoir: at fixed $(T,V,N)$, $F$ is minimized; at fixed $(T,p,N)$, $G$ is minimized. The Legendre structure also encodes stability: convexity of $U$ in extensive variables (equivalently concavity of $S$) forces $C_V\ge 0$ and isothermal compressibility $\kappa_T\ge 0$. Loss of convexity signals a phase transition, handled by the Maxwell equal-area construction.

The microscopic bridge to statistical mechanics: $F=-k_B T\ln Z$ with the canonical partition function $Z=\sum_i e^{-E_i/k_B T}$, $\Omega=-k_B T\ln\Xi$ (grand canonical), and the Gibbs entropy $S = -k_B\sum_i p_i\ln p_i,$ which reduces to Boltzmann's $S = k_B\ln W$ for an equiprobable (microcanonical) ensemble.

4. The Carnot bound and engines

ESTABLISHED A reversible engine operating between reservoirs at $T_h>T_c$ attains the maximum possible efficiency, $\eta_{\rm Carnot} = 1 - \frac{T_c}{T_h},$ and no engine between the same two reservoirs can exceed it (Carnot's theorem, a direct corollary of the second law). This relation defines the thermodynamic (absolute) temperature scale via $T_c/T_h \equiv Q_c/Q_h$ for reversible operation, independent of working substance. Reversibility demands quasi-static operation and hence vanishing power. Finite-power optimization for endoreversible engines yields the Curzon–Ahlborn estimate $\eta_{CA}=1-\sqrt{T_c/T_h}$ INFERENCE — a useful benchmark for a particular class of models, not a universal law.

5. Arrow of time and the statistical reframing

ESTABLISHED The microscopic laws — Hamiltonian mechanics, the Schrödinger/von Neumann equation — are time-reversal invariant (modulo weak CP-violation, irrelevant here), yet macroscopic entropy increase selects a direction of time. Boltzmann's $H$-theorem shows that under the Stosszahlansatz (molecular-chaos assumption) the kinetic-theory functional obeys $dH/dt\le 0$, so $S_B=-k_B H$ increases. INFERENCE; interpretation [CONTESTED] Reconciling microscopic reversibility with macroscopic irreversibility rests on (a) a low-entropy initial condition (the "Past Hypothesis") and (b) the overwhelming phase-space volume of high-entropy macrostates. The arrow is thus statistical, not mechanical — a point developed further on cosmology.

6. Fluctuation theorems — the second law as an equality

ESTABLISHED For a system driven by an external protocol $\lambda(t)$ from an initial canonical equilibrium, the Jarzynski equality holds exactly, arbitrarily far from equilibrium: $\big\langle e^{-\beta W}\big\rangle = e^{-\beta\,\Delta F},\qquad \beta = 1/k_B T,$ where $W$ is the stochastic work over a single realization and $\Delta F$ the equilibrium free-energy difference between endpoints. By Jensen's inequality $\langle e^{-\beta W}\rangle \ge e^{-\beta\langle W\rangle}$, this implies the second law as an average, $\langle W\rangle \ge \Delta F$, while the equality captures the rare trajectories that transiently "violate" it. The finer Crooks fluctuation theorem, $\frac{P_F(+W)}{P_R(-W)} = e^{\beta(W-\Delta F)},$ relates the forward and time-reversed work distributions; Jarzynski is its integral consequence. The general detailed fluctuation theorem for total entropy production $\Sigma$ over time $\tau$ reads $P(+\Sigma)/P(-\Sigma)=e^{\Sigma/k_B}$, with integral form $\langle e^{-\Sigma/k_B}\rangle = 1$ and hence $\langle\Sigma\rangle\ge 0$. ESTABLISHED These relations have been verified experimentally (single-molecule pulling, colloidal beads in optical traps, electronic circuits). Stochastic thermodynamics extends $\delta Q$, $W$, and $S$ to individual trajectories via Langevin and master-equation dynamics with local detailed balance.

7. Landauer's principle (information thermodynamics)

ESTABLISHED as a theorem within the stochastic/statistical framework; [INFERENCE on the universality of its physical inevitability] Erasing one bit of information in contact with a bath at temperature $T$ dissipates at least $W_{\rm erase} \ge k_B T \ln 2,$ equivalently $\Delta S_{\rm env}\ge k_B\ln 2$ per bit. It is logical irreversibility — a 2-to-1 logical map collapsing accessible phase-space volume by a factor of two — that carries the cost. This resolves the Maxwell's-demon paradox (Szilard, Landauer, Bennett): the demon's measurement record must eventually be reset, and that erasure pays the entropy bill. Reversible computation in principle costs nothing. The principle has been confirmed in colloidal and nanomagnetic single-bit experiments. The deep structural identity with Shannon information is developed on information theory.

8. Black-hole thermodynamics

[ESTABLISHED within semiclassical gravity (QFT on curved spacetime); the underlying microstate counting is OPEN] A stationary black hole obeys four laws isomorphic to the laws of thermodynamics (Bardeen–Carter–Hawking):

• 0th: surface gravity $\kappa$ is constant over the horizon (analog of uniform $T$).
• 1st: $dM = \dfrac{\kappa}{8\pi G}\,dA + \Omega_H\,dJ + \Phi\,dQ$.
• 2nd: classically $dA\ge 0$ (Hawking's area theorem).

Bekenstein conjectured, and Hawking's derivation of thermal emission confirmed, the identifications $S_{\rm BH} = \frac{k_B c^3 A}{4 G\hbar} = \frac{A}{4\ell_P^2}\,k_B,\qquad T_H = \frac{\hbar c^3}{8\pi G k_B M} = \frac{\hbar\kappa}{2\pi k_B c},$ the latter Hawking temperature quoted for Schwarzschild. The generalized second law (GSL) posits that $S_{\rm gen}=S_{\rm BH}+S_{\rm out}$ is non-decreasing, $dS_{\rm gen}\ge 0$, rescuing the ordinary second law when matter falls through the horizon. That entropy scales with horizon area rather than enclosed volume is the seed of the holographic principle and the covariant (Bousso) entropy bound $S\le A/4\ell_P^2$. See general relativity and quantum field theory.

---

Foundational assumptions

---

Domain of validity

ESTABLISHED Classical equilibrium thermodynamics is quantitatively exact for systems satisfying all of:

1. Macroscopic size — $N\sim N_A$ degrees of freedom, so relative fluctuations $\sim N^{-1/2}$ are negligible.
2. Short-range interactions — so energy and entropy are extensive and the thermodynamic limit exists.
3. Equilibrium or near-equilibrium — state variables are well-defined (local equilibrium for irreversible thermodynamics à la Onsager and de Groot–Mazur).
4. Separation of timescales — observation times long compared to microscopic relaxation but short compared to any slow drift.

Within this regime, equations of state, Maxwell relations, the Carnot bound, and phase-coexistence conditions are exact constraints.

The framework extends, with modifications, into: small/mesoscopic systems and single molecules, where the laws become statistical and the fluctuation theorems carry the exact content; nonequilibrium steady states via stochastic thermodynamics and linear response (Onsager reciprocity, Green–Kubo) near equilibrium; and quantum regimes (quantum thermodynamics), where work becomes a two-point-measurement observable and coherence/entanglement modify the resource accounting. It is reinterpreted at the foundations by statistical mechanics: the deterministic laws are the $N\to\infty$, fluctuation-suppressed shadows of probabilistic ones. Boltzmann's constant sets the scale at which fluctuations matter; once $k_B T$ is comparable to characteristic energies and $N$ is small, thermodynamic certainties dissolve into distributions.

---

Where it breaks down

• Small systems and single molecules ESTABLISHED — domain boundary, not inconsistency. With $N^{-1/2}$ fluctuations no longer negligible, the second-law inequality is routinely and transiently violated along individual trajectories. "Entropy never decreases" is false at this scale; the fluctuation theorems restore an exact statement only at the level of trajectory ensembles.
• Long-range / self-gravitating systems (stars, galaxies, black holes) ESTABLISHED. Energy and entropy are non-additive and non-extensive; the thermodynamic limit fails. Ensembles become inequivalent (microcanonical $\neq$ canonical), heat capacities can be negative, and black-hole entropy scales as area, not volume. Euler and Gibbs–Duhem do not apply; this forces holographic ideas and is only partially understood. See general relativity.
• Black-hole evaporation and the information paradox OPEN/CONTESTED. Hawking's semiclassical calculation yields exactly thermal radiation, suggesting a pure state evolves to a mixed one — apparent loss of unitarity and of any microstate count behind $S_{\rm BH}$. Whether information is preserved (now widely favored, supported by AdS/CFT, the Page curve, and island/replica-wormhole computations INFERENCE) or lost lacks an agreed bulk mechanism for our (asymptotically flat) universe. The clash is between thermodynamics + quantum unitarity and semiclassical gravity. Tracked in gaps and contradictions.
• Microscopic origin of the arrow of time OPEN. Time-reversal-invariant microdynamics cannot alone produce monotone entropy increase; the $H$-theorem smuggles in the time-asymmetric Stosszahlansatz, and the global arrow requires the cosmological Past Hypothesis. A genuine unsolved foundational gap, not an inconsistency — it exports to cosmology.
• Third law with residual entropy / glasses ESTABLISHED — edge cases. For macroscopic ground-state degeneracy (ice "residual entropy", spin ices) or non-ergodic glasses frozen out of equilibrium, $S$ does not approach zero and equilibrium is operationally unreachable. The Nernst statement holds for true equilibrium with a unique ground state; apparent violations are degeneracy or broken ergodicity — domain mismatches, not failures of quantum statistics.
• Negative absolute temperature ESTABLISHED; interpretation [CONTESTED]. Bounded-spectrum systems reach population-inverted states with $\partial S/\partial U<0$, hence $T<0$, hotter than any positive temperature. Fully consistent thermodynamics, but it breaks the naive "$T>0$"/unbounded-energy presumption. A live dispute (Dunkel–Hilbert vs. proponents of the Boltzmann surface entropy) concerns whether the Gibbs (volume) or Boltzmann (surface) entropy is the correct definition, and thus whether negative temperatures are admissible at all.
• Maxwell's demon / measurement-feedback loops ESTABLISHED resolution. A naive demon appears to violate the second law by sorting molecules; the resolution (Szilard, Landauer, Bennett) charges $k_B T\ln 2$ per erased bit. Not a breakdown but a demonstration that the accounting must include the information-bearing degrees of freedom — the law is sharpened, not refuted.
• Far-from-equilibrium, turbulent, or strongly driven systems OPEN. With steep gradients or strong driving, local equilibrium fails, temperature/entropy may not be definable, and there is no complete predictive macroscopic theory analogous to equilibrium thermodynamics. Extended irreversible thermodynamics, MaxEnt, and large-deviation theory are partial frameworks; a general entropy-production extremum principle (minimum or maximum) is not an established law.

---

Open problems (internal)

• First-principles thermalization and the equal-a-priori-probability postulate. OPEN Ergodic theory, dynamical typicality, and ETH are strong partial results INFERENCE, but a general theorem for generic interacting Hamiltonians does not exist, and many-body-localized systems are explicit non-thermalizing counterexamples. See open problems.
• Origin and magnitude of the cosmological low-entropy initial condition. OPEN A low-entropy past is broadly accepted as logically necessary INFERENCE, but why the early universe had such low gravitational entropy has no accepted explanation (Penrose's Weyl-curvature hypothesis is one SPECULATIVE proposal); it is entangled with quantum gravity.
• Microstate identification behind $S_{\rm BH}$ for realistic black holes. OPEN String theory reproduces $S=A/4$ by counting D-brane microstates for certain supersymmetric/extremal black holes ESTABLISHED within that framework — Strominger–Vafa, and AdS/CFT gives a holographic account INFERENCE, but a universally accepted derivation for generic astrophysical (e.g. Schwarzschild) black holes is lacking.
• An explicit, agreed unitarity-preserving mechanism for black-hole evaporation in asymptotically flat spacetime. OPEN/CONTESTED Island/replica-wormhole computations reproduce the unitary Page curve INFERENCE, within semiclassical-gravity-plus-replica assumptions, strongly favoring information preservation, but the bulk physical mechanism (firewalls? complementarity? new physics?) is unsettled.
• A predictive macroscopic theory for far-from-equilibrium systems. OPEN Linear response and large-deviation theory work near equilibrium; far from it, proposed variational principles are not established as laws.
• The correct entropy/temperature for small and bounded-spectrum systems (Boltzmann surface vs. Gibbs volume entropy). CONTESTED The dispute bears on the precise meaning of $T$ and $S$ away from the thermodynamic limit.
• Foundations of quantum thermodynamics. OPEN Resource-theoretic and two-point-measurement frameworks are internally consistent, but which definition of "work" is physically privileged, and how coherence functions as a thermodynamic resource, are active research questions. See quantum mechanics.

---

Connections to other frameworks

• Statistical mechanics — the foundational substrate. It derives the state functions ($S=k_B\ln W$, $F=-k_B T\ln Z$) and reinterprets the second law as probabilistic. Thermodynamics is the $N\to\infty$, fluctuation-suppressed limit; the two agree wherever they overlap, with tension only over the interpretation of irreversibility and over small-$N$ regimes.
• Quantum mechanics / quantum field theory — discrete spectra underwrite the third law and the microstate count; the von Neumann entropy $S=-k_B\,\mathrm{Tr}(\rho\ln\rho)$ generalizes the Gibbs entropy. Quantum thermodynamics adds coherence and entanglement as resources. Clash points: the definition of work, measurement back-action, and whether unitarity is compatible with thermal (Hawking) radiation.
• General relativity — black-hole mechanics maps exactly onto the four laws, with $S=A/4$ and Hawking temperature; this both validates thermodynamics gravitationally and exports it (GSL, holographic and Bousso bounds, Jacobson's SPECULATIVE derivation of the Einstein equations as an equation of state). The deepest clash — the information paradox — sits at the GR/QM/thermodynamics triple point.
• Cosmology — supplies the low-entropy boundary condition (Past Hypothesis) that thermodynamics presupposes but cannot itself explain; the global entropy budget is dominated by gravitational/black-hole entropy, and the heat-death endpoint is a thermodynamic extrapolation INFERENCE.
• Information theory — the Shannon entropy $H=-\sum p\log p$ is formally identical to the Gibbs entropy up to $k_B$ and the logarithm base; Landauer's principle fixes the physical exchange rate ($k_B T\ln 2$ per bit). Maxwell's demon, Szilard engines, and Sagawa–Ueda feedback thermodynamics unify the two ledgers. A deep structural identity, not a mere analogy.
• Classical mechanics and kinetic theory — Hamiltonian dynamics plus the Boltzmann equation, Onsager reciprocal relations, and Green–Kubo formulas extend equilibrium thermodynamics to transport and the near-equilibrium frontier; stochastic thermodynamics carries the laws into the driven mesoscale.
• Mathematics — convex analysis and the Legendre transform supply the rigorous backbone of the potential structure, stability, and phase transitions.

---

Key references

• H. B. Callen, Thermodynamics and an Introduction to Thermostatistics, 2nd ed., Wiley (1985) — the canonical postulatory treatment (fundamental relation, extremum principles, Legendre transforms, Maxwell relations, stability).
• E. Fermi, Thermodynamics, Dover (1956); M. W. Zemansky & R. H. Dittman, Heat and Thermodynamics — classic expositions of the four laws and absolute temperature.
• L. D. Landau & E. M. Lifshitz, Statistical Physics, Part 1 (Vol. 5), Pergamon — the bridge from statistical mechanics, including fluctuations and the quantum-statistical third law.
• C. Jarzynski, "Nonequilibrium equality for free energy differences," Phys. Rev. Lett. 78, 2690 (1997).
• G. E. Crooks, "Entropy production fluctuation theorem and the nonequilibrium work relation for free energy differences," Phys. Rev. E 60, 2721 (1999).
• U. Seifert, "Stochastic thermodynamics, fluctuation theorems and molecular machines," Rep. Prog. Phys. 75, 126001 (2012).
• R. Landauer, "Irreversibility and heat generation in the computing process," IBM J. Res. Dev. 5, 183 (1961); C. H. Bennett, "The thermodynamics of computation—a review," Int. J. Theor. Phys. 21, 905 (1982).
• J. D. Bekenstein, "Black holes and entropy," Phys. Rev. D 7, 2333 (1973); S. W. Hawking, "Particle creation by black holes," Commun. Math. Phys. 43, 199 (1975).
• J. M. Bardeen, B. Carter & S. W. Hawking, "The four laws of black hole mechanics," Commun. Math. Phys. 31, 161 (1973).
• A. Strominger & C. Vafa, "Microscopic origin of the Bekenstein–Hawking entropy," Phys. Lett. B 379, 99 (1996).
• J. L. Lebowitz, "Boltzmann's entropy and time's arrow," Physics Today 46(9), 32 (1993); L. Boltzmann, Lectures on Gas Theory (1896–98).
• S. R. de Groot & P. Mazur, Non-Equilibrium Thermodynamics, Dover (1984); L. Onsager, Phys. Rev. 37, 405 and 38, 2265 (1931).
• S. Carnot, Réflexions sur la puissance motrice du feu (1824); R. Clausius (1865, introduction of entropy).

---

See also

• Statistical mechanics — microscopic foundations of entropy and the partition function.
• Information theory — Shannon entropy, Landauer's principle, Maxwell's demon.
• General relativity — black-hole mechanics and horizon thermodynamics.
• Quantum mechanics / Quantum field theory — discrete spectra, von Neumann entropy, quantum thermodynamics.
• Cosmology — the Past Hypothesis and the global entropy budget.
• Classical mechanics — Hamiltonian substrate and kinetic theory.
• Mathematics — convex analysis and the Legendre transform.
• THEORY_MAP.md · OPEN_PROBLEMS.md · GAPS_AND_CONTRADICTIONS.md · UNIFYING_PRINCIPLES.md · ASSUMPTIONS_LEDGER.md · CONSTANTS_AND_SCALES.md · GLOSSARY.md · EPISTEMICS.md

References

See BIBLIOGRAPHY.md for the consolidated, cross-page reference list.