The Clausius Inequality and Entropy as a State Function

Derivation

The Clausius Inequality and Entropy as a State Function

D-150 Home PU-203 Threads energy · chance Depends on Carnot's Theorem and Maximum Engine Efficiency

Statement

For any cyclic process a system undergoes while exchanging heat with reservoirs at absolute temperature T, the Clausius integral obeys ∮ dQ/T ≤ 0, with equality if and only if the cycle is reversible. The vanishing of ∮ dQ_rev/T makes ∫ dQ_rev/T path-independent, so dS = dQ_rev/T defines a state function, the entropy S; and for any real process dS ≥ dQ/T, whence the entropy of an isolated system never decreases.

Why it matters

This is the theorem that turns the second law from a list of impossible engines into a single scalar state function. Everything downstream — chemical potentials, phase equilibria, the Gibbs and Helmholtz free energies, the arrow of time — hangs on the fact that dQ_rev/T is an exact differential while dQ alone is not.

It is also the precise sense in which "energy is conserved but availability is not." The first law counts joules; the Clausius inequality tells you how many of those joules have been degraded into forms from which no further work can be extracted, and it does so with a bookkeeping quantity, S, that depends only on the state and not on how you got there.

Assumptions

The Kelvin–Planck statement of the second law holds.No cyclic device may convert heat from a single reservoir wholly into work. Drop it and the auxiliary Carnot engines could pump net work out of the master reservoir, the sign of the inequality is lost, and ∮ dQ/T could take either sign. A thermodynamic (absolute) temperature scale exists, with T > 0.Established by Carnot efficiency: a reversible engine between two reservoirs satisfies Q_H/Q_C = T_H/T_C. Without it the ratio dQ/T has no reservoir-independent meaning and the integrand is undefined. The system returns exactly to its initial state each cycle, and the auxiliary Carnot engines can be made arbitrarily close to reversible.If the cycle does not close, ΔU ≠ 0 and the first-law bookkeeping that equates net work to master-reservoir heat fails; if the auxiliary engines carry their own irreversibility they inject extra entropy and the equality branch is never reached. At every contact the system's boundary temperature equals the reservoir temperature T supplying dQ.For an irreversible cycle the relevant T is the reservoir (boundary) temperature, not the system's ill-defined internal temperature; conflating the two mis-locates the inequality.

Derivation

a cycle 𝒞 : divide into N steps, step i exchanging dQ_i with a reservoir at T_i

Any smooth cyclic path may be approximated to arbitrary accuracy by a finite chain of isothermal contacts; the continuum limit is taken at the end. A

insert a reversible Carnot engine between a master reservoir T₀ and each T_i, feeding dQ_i into the system

Legal because a Carnot engine can deliver any required heat at any T_i while touching only T₀ and T_i; this replaces every reservoir in the problem with the single master reservoir T₀. B

dQ_0,i = T₀ · (dQ_i / T_i)

The thermodynamic-temperature relation for a reversible engine, Q_hot/Q_cold = T_hot/T_cold, applied to each auxiliary engine (heat drawn from T₀ to supply dQ_i at T_i). This is the only place carnot-efficiency enters. B

Q₀ = Σ_i dQ_0,i = T₀ Σ_i dQ_i/T_i ⟶ T₀ ∮_𝒞 dQ/T

Total heat withdrawn from the master reservoir over one full pass; T₀ is a constant and factors out, and N → ∞ turns the sum into the closed-loop integral. A

system + N auxiliary engines all return to their initial states ⟹ ΔU_total = 0 ⟹ W_net = Q₀

First law over one complete cycle of the composite device: with internal energy restored, every joule taken from T₀ emerges as net work. B

Kelvin–Planck: W_net ≤ 0 ⟹ Q₀ = T₀ ∮ dQ/T ≤ 0 ⟹ ∮ dQ/T ≤ 0

The composite exchanges heat with the single reservoir T₀ over a cycle; the second law forbids net positive work from that. Dividing by T₀ > 0 preserves the sense of the inequality. This is the Clausius theorem. C

for a reversible 𝒞 run backward every dQ → −dQ, giving ∮ dQ/T ≥ 0; with step 6: ∮ dQ_rev/T = 0

A reversible cycle satisfies the inequality in both directions, and the only number that is both ≤ 0 and ≥ 0 is zero. C

∮ = ∫_{A→B (path 1)} + ∫_{B→A (path 2)} = 0 ⟹ ∫_{A→B (path 1)} dQ_rev/T = ∫_{A→B (path 2)} dQ_rev/T

Two arbitrary reversible paths from A to B form a closed loop; the loop integral vanishes, so the two path integrals are equal. Path-independence of the integral is exactly the condition for an exact differential. B

define dS ≡ dQ_rev/T , S_B − S_A = ∫_A^B dQ_rev/T

Because the integral depends only on the endpoints, it defines a state function up to an additive constant — the entropy. A

irreversible A→B, then reversible B→A: ∫_A^B dQ/T + ∫_B^A dQ_rev/T ≤ 0

This composite is a genuine cycle, so the Clausius theorem (step 6, strict for an irreversible leg) applies. C

∫_B^A dQ_rev/T = S_A − S_B ⟹ ∫_A^B dQ/T ≤ S_B − S_A ⟹ dS ≥ dQ/T

Substitute the state-function value of the reversible return leg and rearrange; in differential form this is the Clausius inequality for a process. C

isolated system: dQ = 0 ⟹ dS ≥ 0

With no heat crossing the boundary the inequality collapses to the law of entropy increase. A

Result

∮ dQ/T ≤ 0 (= 0 iff reversible) ⟹ dS = dQ_rev/T , dS ≥ dQ/T , (dS)_isolated ≥ 0

Reading. Weighting each heat exchange by the reciprocal of the temperature at which it happens produces a quantity that returns to zero over any reversible cycle. That closure is what promotes S to a state variable: its change between two states is fixed regardless of path, even a wildly irreversible one, because you may compute it along any convenient reversible substitute. For real processes the inequality records the shortfall — entropy is generated, never destroyed.

Units check. dQ is in joules (J) and T in kelvin (K), so dQ/T and S carry units of J·K⁻¹. The loop integral ∮ dQ/T is a sum of J·K⁻¹ terms, dimensionally consistent with an entropy change, and the dimensionless inequality ≤ 0 compares like with like.

Limiting cases

Reversible limit: all inequalities become equalities; ∮ dQ_rev/T = 0 and dS = dQ_rev/T exactly — no entropy generated.
Adiabatic reversible (isentropic): dQ = 0 and reversible ⟹ dS = 0; the process runs along a surface of constant entropy.
Adiabatic irreversible: dQ = 0 but dS > 0 — e.g. free expansion; entropy rises with no heat flow at all.
Single-reservoir isothermal cycle: ∮ dQ/T = (1/T)∮ dQ = W/T ≤ 0, recovering Kelvin–Planck directly.
Carnot cycle: two isotherms give Q_H/T_H − Q_C/T_C = 0, the discrete form of the equality.

Breaks when

Negative absolute temperatures. In population-inverted spin systems T < 0 on the Kelvin scale; dividing by a negative T flips the inequality's sense, and the Carnot-chaining proof (which needs T₀ > 0 to preserve the sign in step 6) no longer delivers ∮ dQ/T ≤ 0. Entropy is still defined, but its ordering with temperature is inverted.
Small systems / large fluctuations. For a handful of particles the heat exchanged in a cycle fluctuates and ∮ dQ/T can be transiently positive on individual realisations. The Clausius inequality survives only as a statement about the ensemble average, ⟨∮ dQ/T⟩ ≤ 0; the sharp per-trajectory bound is replaced by fluctuation theorems (Jarzynski, Crooks).
No well-defined boundary temperature. If the system is driven so violently that the reservoir contact temperature is ambiguous (shock fronts, radiative non-equilibrium), the integrand dQ/T has no single T to use and the inequality cannot be evaluated as written.
Long-range / self-gravitating systems. Where entropy is non-extensive and no proper equilibrium reservoir exists (self-gravitating gas, black-hole thermodynamics without the horizon prescription), the reservoir-chaining construction has no valid T₀ and the classical proof does not apply.

Failure modes

Using the system temperature instead of the boundary temperature in dQ/T for an irreversible step. The Clausius T is the reservoir temperature at the contact; the system may not even have a uniform T.
Computing ΔS along the actual irreversible path. S is a state function — integrate dQ_rev/T along a reversible substitute joining the same endpoints, never the real dissipative path.
Claiming ΔS = 0 for an adiabatic free expansion because "no heat flowed." Q = 0 forces dS ≥ 0, not dS = 0; the expansion is irreversible so ΔS > 0.
Forgetting the surroundings. A system's entropy can fall (refrigeration); only the total (system + reservoirs) obeys ΔS_total ≥ 0.
Sign-convention slips in dQ — treating rejected heat as positive — which spuriously flip the direction of the inequality.
Treating ∮ dQ/T ≤ 0 as a first-law statement. It is not energy balance; ∮ dQ equals the net work, whereas ∮ dQ/T is a second-law quantity that is generally nonzero even when energy is conserved.

Discussion

The deepest content of the theorem is the promotion of a path-dependent quantity to a state function. Heat dQ is an inexact differential: its cyclic integral, the net work, is generally nonzero, which is exactly why an engine can run. The integrating factor 1/T — and the fact that the thermodynamic temperature is a valid integrating factor is the real theorem — converts dQ_rev into the exact differential dS. That Carnot's efficiency result supplies precisely this factor is the hinge on which the whole edifice turns.

Physically, entropy generation measures lost opportunity. Whenever heat crosses a finite temperature difference, work that a reversible engine could have extracted is thrown away, and the ledger of that loss is ΔS_gen = ΔS_total ≥ 0. The inequality is therefore not merely a prohibition but a currency: it quantifies how far a real process falls short of the reversible ideal, and it is additive across independent subsystems, which is why it scales up to whole plants and down to single molecular motors.

The theorem also fixes the direction of time in an otherwise time-symmetric mechanics. Newton's and Maxwell's equations run equally well forward and backward; the Clausius inequality does not. Its origin is statistical — the overwhelming number of microstates consistent with a "mixed" macrostate — and Boltzmann's S = k_B ln Ω later gave the state function a microscopic identity, with k_B carrying the same J·K⁻¹ units that the Clausius integral demanded.

At the frontier the sharp inequality softens into an equality once fluctuations are tracked. The Jarzynski relation ⟨e^−W/k_BT⟩ = e^−ΔF/k_BT and the Crooks fluctuation theorem reproduce ⟨W⟩ ≥ ΔF (equivalently ΔS_total ≥ 0) on averaging, while permitting individual trajectories to transiently violate it. In this light the Clausius inequality is the macroscopic, thermodynamic-limit shadow of an exact microscopic identity — the second law is not so much broken as revealed to be a statement about ensembles.

Common misconceptions. Entropy is not "disorder" in any everyday visual sense — a crystallising solution can lower its own entropy while raising the surroundings' more; it is not the same thing as heat, nor is it conserved; and dS ≥ dQ/T does not forbid a system's entropy from decreasing, only the total's.

Worked examples

Example 1 — The Clausius integral vanishes for a Carnot cycle.

cycle = isotherm at T_H (absorb Q_H) + adiabat + isotherm at T_C (reject Q_C) + adiabat

Adiabatic legs carry dQ = 0 and contribute nothing to ∮ dQ/T. A

∮ dQ/T = Q_H/T_H − Q_C/T_C

Each isotherm has constant T, so ∫ dQ/T = Q/T; rejected heat enters with a minus sign. A

carnot-efficiency: Q_H/Q_C = T_H/T_C. Take T_H = 500 K, T_C = 300 K, Q_H = 1000 J ⟹ Q_C = 600 J

Thermodynamic-temperature relation for the reversible engine; then insert numbers. B

∮ dQ/T = 1000/500 − 600/300 = 2.00 − 2.00 J·K⁻¹

Direct arithmetic. A

∮ dQ/T = 0.00 J·K⁻¹

Reading. The reversible cycle closes the entropy ledger exactly; the two isothermal contributions cancel to the last figure. Units check. J divided by K gives J·K⁻¹, and the difference of two equal J·K⁻¹ quantities is zero.

Example 2 — Entropy generated by heat flow across a finite gradient.

Q flows from reservoir T_H to reservoir T_C; each reservoir stays at fixed T (very large heat capacity)

A reservoir's temperature is unchanged by finite Q, so its ΔS = ∫ dQ_rev/T = Q/T at that fixed T. A

ΔS_total = ΔS_cold + ΔS_hot = (+Q/T_C) + (−Q/T_H) = Q(1/T_C − 1/T_H)

Cold gains +Q, hot loses Q; entropy is a state function so we sum the reservoir changes. B

T_H = 400 K, T_C = 300 K, Q = 1000 J

Insert numbers after the symbolic form is fixed. A

ΔS_total = 1000(1/300 − 1/400) = 1000(3.333×10⁻³ − 2.500×10⁻³) = 1000 × 8.33×10⁻⁴

Arithmetic; the bracket is positive because T_C < T_H. A

ΔS_total = +0.833 J·K⁻¹ > 0

Reading. Spontaneous heat flow down a temperature gradient generates entropy; the process is irreversible, consistent with dS > dQ/T for the composite. Reversing it (heat flowing cold→hot unaided) would give ΔS_total < 0 and is forbidden. Units check. J × K⁻¹ = J·K⁻¹, the correct entropy unit.

Problems

A reversible engine takes Q_H = 2400 J from a reservoir at 600 K and rejects heat to a reservoir at 300 K. Verify that ∮ dQ/T = 0.
Solution
By carnot-efficiency, Q_C/Q_H = T_C/T_H = 300/600 = 0.5, so Q_C = 1200 J. Then ∮ dQ/T = Q_H/T_H − Q_C/T_C = 2400/600 − 1200/300 = 4.00 − 4.00 = 0 J·K⁻¹. The reversible cycle closes the entropy ledger.
One mole of ideal gas expands isothermally and reversibly at T = 300 K from V₁ to V₂ = 2V₁. Find ΔS of the gas, and state ΔS for the same endpoints reached by an irreversible free expansion.
Solution
Isothermal ideal gas: dU = 0, so dQ_rev = dW = p\,dV = (nRT/V)dV. Thus ΔS = ∫ dQ_rev/T = nR ∫_V₁^V₂ dV/V = nR ln(V₂/V₁) = (1)(8.314) ln 2 = 5.76 J·K⁻¹. Because S is a state function, the free expansion between the same two states has the identical ΔS_gas = +5.76 J·K⁻¹ — but there Q = 0, so entropy is generated rather than transferred, and ΔS_total = +5.76 J·K⁻¹ too (surroundings unchanged).
Two identical copper blocks, heat capacity C = 400 J·K⁻¹ each, start at 350 K and 250 K and are placed in thermal contact until they equilibrate. Find the final temperature and the total entropy generated.
Solution
Equal heat capacities ⟹ T_f = (350 + 250)/2 = 300 K. For each block ΔS = C ln(T_f/T_i) (integrating dQ_rev/T = C\,dT/T along a reversible warming/cooling substitute). Hot block: 400 ln(300/350) = 400(−0.15415) = −61.66 J·K⁻¹. Cold block: 400 ln(300/250) = 400(+0.18232) = +72.93 J·K⁻¹. Total ΔS_gen = −61.66 + 72.93 = +11.3 J·K⁻¹ > 0, confirming irreversibility.
Ice melts at 0 °C (273.15 K) absorbing latent heat L = 334 J·g⁻¹. For 50 g melted by contact with surroundings at 20 °C (293.15 K), find the entropy change of the ice and of the total system.
Solution
Melting is isothermal at T_m = 273.15 K. Heat absorbed Q = mL = 50 × 334 = 16700 J. Ice: ΔS_ice = Q/T_m = 16700/273.15 = +61.14 J·K⁻¹. Surroundings lose Q at 293.15 K: ΔS_surr = −16700/293.15 = −56.97 J·K⁻¹. Total ΔS_total = 61.14 − 56.97 = +4.17 J·K⁻¹ > 0 — the finite temperature gap makes it irreversible.
Using only the Clausius inequality, prove that no engine operating between two reservoirs T_H > T_C can exceed the Carnot efficiency η_C = 1 − T_C/T_H.
Solution
For any engine cycle, ∮ dQ/T ≤ 0 gives Q_H/T_H − Q_C/T_C ≤ 0 (heats as magnitudes, rejected heat negative). Hence Q_C/Q_H ≥ T_C/T_H. Efficiency η = W/Q_H = 1 − Q_C/Q_H ≤ 1 − T_C/T_H = η_C. Equality holds only for a reversible (Carnot) cycle, where ∮ dQ/T = 0. Thus the Clausius inequality directly bounds every engine by the Carnot value.