Thermodynamics
Entropy and the Second Law
dS_universe ≥ 0 — heat flows hot→cold spontaneously; Boltzmann S = k_B ln W relates micro- to macrostate
Entropy is a state function with units of joules per kelvin that quantifies how a system's energy is spread among its accessible configurations. The second law of thermodynamics states that the total entropy of an isolated system never decreases — dSuniverse ≥ 0 for any process, with equality only at reversible limit. Macroscopically, dS = δqrev/T (Clausius 1854); microscopically, S = kB ln W (Boltzmann 1877, Planck 1900) where W is the number of microscopic configurations and kB = 1.380649 × 10⁻²³ J/K. Heat flows from hot to cold without external help; the reverse requires work input — exactly the asymmetry that lets engines do useful work while capping their efficiency at ηCarnot = 1 − Tc/Th.
- Boltzmann constantk_B = 1.380649 × 10⁻²³ J/K
- MacroscopicdS = δq_rev/T
- MicroscopicS = k_B ln W
- Carnot ceilingη = 1 − T_c/T_h
- Argon (298 K)154.84 J/(K·mol)
- StatedClausius 1854; Kelvin 1851; Boltzmann 1872
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why entropy and the second law matter
- Caps every heat engine. A coal plant with superheated steam at 833 K rejecting to a 308 K cooling tower has Carnot ceiling η = 1 − 308/833 = 63%; real plants achieve 38-45% because of irreversibilities (friction, finite-rate heat transfer, non-ideal expansion). Combined-cycle gas turbines push this to ~60% by stacking two Brayton/Rankine cycles between 1700 K and 300 K.
- Defines absolute temperature. Two systems in thermal contact equilibrate by maximizing total entropy, which forces 1/T = (∂S/∂E)V,N. This is more fundamental than a thermometer reading — it makes T a derivative of a counting function, not just a property of mercury column length.
- Sets equilibrium conditions for chemistry. At constant T and P, the criterion of spontaneity is dG < 0 where G = H − TS. Entropy decides whether endothermic processes proceed: NH₄NO₃ dissolution is endothermic (ΔH = +25.7 kJ/mol) but proceeds because ΔS = +108.7 J/(K·mol), giving ΔG = −6.7 kJ/mol at 298 K. Without the entropy term most life-relevant reactions would be impossible.
- Quantifies information. Shannon's 1948 information entropy H = −Σ p_i log p_i has the same mathematical form as Gibbs's 1878 thermodynamic entropy. Landauer's 1961 principle ties them: erasing one bit at temperature T dissipates at least k_B T ln 2 ≈ 2.85 zJ at 298 K. Modern data centers approach this 'Landauer limit' as a fundamental floor.
- Explains the arrow of time. Microscopic dynamics are time-reversal symmetric (Hamilton's equations, Schrödinger equation), but macroscopic processes are unidirectional because the high-entropy macrostate is overwhelmingly more probable. Boltzmann's H-theorem (1872) was the first proof that a low-entropy initial state evolves toward equilibrium under molecular collisions.
- Heat death of the universe. If general relativity plus thermodynamics extrapolate, the cosmos approaches a state of maximum entropy where no usable energy gradients remain — Helmholtz 1854. Modern cosmology refines this with cosmological event horizons; the Bekenstein-Hawking entropy of the observable universe today is ~10¹²² k_B, dominated by the cosmological-horizon area, with stellar/black-hole entropy contributing ~10¹⁰⁴ k_B.
- Refrigeration costs work. Moving Q heat from cold reservoir T_c to hot T_h requires minimum work W ≥ Q·(T_h/T_c − 1). A typical home refrigerator at T_c = 277 K and T_h = 298 K needs W ≥ Q·0.076 — best-in-class units achieve coefficients of performance around 4-5, only 30-40% of the Carnot limit (~13).
Common misconceptions
- "Entropy means disorder." Entropy is the logarithm of microstate multiplicity W. A perfect crystal has S = 0 not because it is "ordered" but because there is exactly one accessible configuration at 0 K. Phase separations (oil floating on water) can be lower entropy than a uniform mixture but appear more "ordered" by eye — the colloquial sense and thermodynamic definition diverge for any system with hidden internal degrees of freedom.
- "The second law is violated by life or by evolution." Open systems can locally decrease entropy at the cost of larger entropy increases in their surroundings. A cell making proteins from amino acids drops local S by ~10⁻²¹ J/K per peptide bond formed, while exporting ~10⁻¹⁹ J/K of metabolic-heat entropy to the medium per ATP hydrolyzed.
- "Entropy always increases." Only entropy of isolated (or non-interacting subsystem of a) total system never decreases. A subsystem in contact with another can lose entropy as long as the partner gains more. The fluctuation theorem (Evans-Searles 1994; Crooks 1999) further says entropy can transiently decrease in small systems with probability exp(−ΔS/k_B), which has been measured directly in colloidal-bead experiments (Wang et al 2002).
- "k_B ln W requires equilibrium." The Boltzmann formula is a definition — it counts accessible microstates whether or not the system is in equilibrium. What requires equilibrium is the equality between Boltzmann entropy and Clausius entropy. Out of equilibrium, multiple competing definitions exist (Gibbs entropy, Kolmogorov-Sinai entropy, von Neumann entropy) and need careful contextual choice.
- "The second law demands strict increase." dS ≥ 0 — equality holds in reversible processes. A quasi-static isothermal expansion of an ideal gas conserves total entropy: gas gains nR ln(V₂/V₁), reservoir loses the same. Real processes have positive entropy production proportional to fluxes squared (linear-response regime) or worse.
- "Maxwell's demon refutes the second law." No — Bennett 1982 showed the demon's memory storage and erasure must dissipate at least k_B T ln 2 per bit, exactly compensating the entropy decrease the demon achieves by sorting. The 2010 Toyabe et al experiment with a feedback-controlled colloidal bead extracted ~10⁻²¹ J of work per measurement, consistent with the Landauer bound.
Derivation
Begin with Carnot's 1824 result that the maximum efficiency of any heat engine cycling between Th and Tc is η = 1 − Tc/Th independent of working substance. Cyclically integrating around any reversible cycle gives ∮ δqrev/T = 0, which means δqrev/T is the differential of a state function. Clausius (1865) named it entropy: dS ≡ δqrev/T. For irreversible processes the Clausius inequality dS ≥ δq/T holds, and for an isolated system δq = 0, recovering dSisolated ≥ 0.
The microscopic side comes from Boltzmann's 1877 counting argument. For an isolated system at fixed energy E, all accessible microstates are equally likely (microcanonical ensemble). The number of microstates consistent with the macrostate is W(E,V,N). Define S = kB ln W. For two systems in thermal contact with shared total energy E = E1 + E2, the joint W = W1(E1)·W2(E−E1) is maximized at the equilibrium energy split, giving (∂ ln W1/∂E1) = (∂ ln W2/∂E2), i.e. (∂S1/∂E1) = (∂S2/∂E2). Defining T by 1/T ≡ (∂S/∂E) recovers thermal equilibrium as equal temperature. The factor kB aligns the J/K units with the Clausius definition; agreement requires the exact value kB = 1.380649 × 10⁻²³ J/K (now a defined SI constant since 2019).
For an ideal monatomic gas, explicit counting in a 6N-dimensional phase space gives the Sackur-Tetrode equation: S = nR[ln((V/n)·(2π·m·kBT/h²)^(3/2)) + 5/2]. At STP for argon (m = 6.634 × 10⁻²⁶ kg, V/n = 24.79 L/mol, T = 298.15 K) this evaluates to S = 154.84 J/(K·mol), agreeing with calorimetric integration of Cp/T from 0 K within experimental uncertainty — the strongest operational test of S = kB ln W.
Three formulations of entropy
| Formulation | Author | Definition | Domain | Strengths |
|---|---|---|---|---|
| Clausius | Clausius 1854/1865 | dS = δq_rev/T | Macroscopic, reversible processes | Calorimetrically measurable; first-law-compatible |
| Boltzmann | Boltzmann 1877; Planck 1900 | S = k_B ln W | Microcanonical ensemble at equilibrium | Connects micro- and macrostates; intuitive counting |
| Gibbs | Gibbs 1878 | S = −k_B Σ p_i ln p_i | Any ensemble, equilibrium or not | Generalizes; reduces to Boltzmann when p_i = 1/W |
| von Neumann | von Neumann 1927 | S = −k_B Tr(ρ ln ρ) | Quantum statistical mechanics | Density-matrix form for entangled systems |
| Shannon | Shannon 1948 | H = −Σ p_i log₂ p_i | Information theory | Same form, base-2 log → bits; ties to Landauer |
| Bekenstein-Hawking | Bekenstein 1972, Hawking 1974 | S = k_B·c³·A/(4Gℏ) | Black hole event horizons | Quarter-area in Planck units |
| Tsallis | Tsallis 1988 | S_q = k_B(1−Σp_i^q)/(q−1) | Long-range / non-extensive systems | q→1 recovers Gibbs; nonadditive otherwise |
Standard molar entropy at 298.15 K (J/(K·mol))
| Substance | State | S° (J/(K·mol)) | Notes |
|---|---|---|---|
| C (graphite) | solid | 5.74 | Reference for ΔH_f° = 0 |
| C (diamond) | solid | 2.38 | Stiffer lattice → fewer phonon modes |
| H₂O | liquid | 69.95 | Hydrogen-bond ordering reduces S |
| H₂O | gas (steam) | 188.84 | +118.9 J/(K·mol) from translation/rotation |
| He | gas | 126.15 | Sackur-Tetrode confirms |
| Ar | gas | 154.84 | Heavier than He; more translational states |
| O₂ | gas | 205.15 | Adds rotational modes |
| CO₂ | gas | 213.79 | Linear triatomic with vibrational modes |
| Glucose | solid | 212.0 | Many internal rotations frozen |
Heat-engine and refrigerator efficiency limits
| Device | T_h (K) | T_c (K) | Carnot η or COP | Real-world value |
|---|---|---|---|---|
| Coal-fired Rankine | 833 | 308 | η = 0.63 | 0.38-0.42 |
| Combined-cycle gas turbine | 1700 | 300 | η = 0.82 | 0.60-0.62 |
| Automobile Otto cycle | ~2500 | ~300 | η = 0.88 | 0.20-0.30 |
| Solar PV (radiative limit) | 5800 (Sun) | 300 (cell) | η = 0.95 | 0.22-0.27 single-junction |
| Home refrigerator | 298 | 277 | COP_C = 13.2 | 3-5 |
| Cryogenic He liquefier | 300 | 4.2 | COP_C = 0.014 | 0.001-0.003 |
Famous experiments and applications
- Joule's mechanical equivalent (1845). James Prescott Joule's paddle-wheel experiment in a calorimeter showed that 4.184 J of mechanical work raises 1 g water by 1 K, fixing the conversion between calorie and joule and underwriting the first law that closed the door on caloric theory. Joule's measurements demanded that energy be conserved as heat — the prerequisite for Clausius and Kelvin's second-law arguments a few years later.
- Boltzmann's H-theorem (1872). Proved using the assumption of molecular chaos (Stosszahlansatz) that an ideal gas's H functional, equivalent to negative entropy, monotonically decreases under collisional dynamics. Formally established irreversibility from underlying reversible mechanics. Loschmidt's 1876 reversal paradox and Zermelo's 1896 recurrence objection both required Boltzmann to clarify the statistical (not absolute) character of the second law.
- Sackur-Tetrode independent derivations (1911-1912). Otto Sackur and Hugo Tetrode separately derived the absolute entropy of a monatomic ideal gas using the new quantum hypothesis: S includes the term ln(h³). Comparison with calorimetrically integrated entropy confirmed Planck's constant and validated the microscopic interpretation of entropy.
- Penzias and Wilson 1965 cosmic microwave background. Detected the 2.725 K thermal radiation predicted by Big Bang cosmology. The CMB is the largest reservoir of entropy in the observable universe (~10⁸⁹ k_B from photons and neutrinos), dwarfing baryonic matter (~10⁸¹ k_B) — direct cosmological-scale evidence that entropy is monotonically increasing as the universe expands and cools.
- Toyabe et al 2010 Maxwell-demon experiment. A 287 nm polystyrene bead in a viscous fluid, subjected to feedback control via real-time imaging, was shown to extract ~10⁻²¹ J of work from thermal fluctuations per measurement — exactly compensating the Landauer cost of the information used (k_B T ln 2 ≈ 2.85 zJ at 297 K). First direct laboratory confirmation that the Maxwell demon does not violate the second law once you include the demon's memory-erasure entropy.
Frequently asked questions
What is entropy in simple terms?
Entropy S is a state function with units of J/K that measures how spread out a system's energy is among its accessible microstates. Macroscopically, dS = δq_rev/T — the heat absorbed reversibly divided by the absolute temperature. Microscopically, S = k_B ln W where W is the number of microscopic configurations consistent with the macroscopic state and k_B = 1.380649 × 10⁻²³ J/K. A monatomic ideal gas at room temperature has S ≈ 154 J/(K·mol) for argon. A perfect crystal at 0 K has S = 0 (third law). Entropy is not 'disorder' in the colloquial sense — it is logarithm of multiplicity, a precise count.
How is the second law actually stated?
Several equivalent forms exist. Clausius (1854): heat cannot spontaneously flow from a colder to a hotter body. Kelvin-Planck (1851): no cyclic process can completely convert heat into work without other changes. Carathéodory (1909): in any neighbourhood of any state, there exist states inaccessible by adiabatic processes. Statistical (Boltzmann): the entropy of an isolated system, defined as k_B ln W, almost never decreases for a macroscopic system. The unifying mathematical content is dS_universe ≥ 0 for any process, with equality only for reversible processes. Practical consequence: the maximum efficiency of a heat engine running between T_h and T_c is η_Carnot = 1 − T_c/T_h.
Why does Boltzmann's S = k_B ln W matter?
It connects the macroscopic Clausius entropy (a calorimetric quantity measured by integrating dq/T) to the microscopic counting of accessible quantum states. For two systems in thermal contact, the equilibrium maximizes total W = W_1·W_2, equivalent to maximizing ln W_1 + ln W_2 = (S_1 + S_2)/k_B. Setting d(S_1 + S_2)/dE = 0 for fixed total energy gives the equality 1/T_1 = 1/T_2 — defining temperature as ∂S/∂E. The formula appears engraved on Boltzmann's tombstone in Vienna's Zentralfriedhof in the form 'S = k log W' (he died 1906; the formula was Planck's compact rewriting of Boltzmann's 1877 work).
Does the second law mean perpetual motion is impossible?
Yes — perpetual motion of the second kind, defined as a cyclic device that converts heat from a single reservoir entirely into work, violates the Kelvin-Planck statement and is impossible. Real heat engines must reject some heat to a colder reservoir, capping efficiency at η_Carnot = 1 − T_c/T_h. A power plant with steam at 833 K and condenser at 308 K has Carnot ceiling 63%, not the 100% naive 'energy in = work out' would suggest. Perpetual motion of the first kind (creates energy) violates the first law (energy conservation) instead. The US Patent Office requires a working model for perpetual-motion applications and has rejected such patents on first principles for over a century.
Doesn't life decrease entropy locally?
Living organisms are open systems, not isolated. They maintain low local entropy by exporting more entropy to their surroundings than they create internally. A 70 kg human at 310 K dissipates ~100 W of metabolic heat to the environment, generating entropy at ~100/310 ≈ 0.32 J/(K·s) — about 28000 J/K per day. Over a 70-year lifetime that is 7 × 10⁸ J/K, far exceeding the ~10⁵ J/K an adult human body's organized structure represents. Schrödinger described this in 'What Is Life?' (1944) as life feeding on negative entropy — more precisely, on an entropy gradient between solar photons (T_eff ≈ 5800 K) and Earth's IR re-emission to space (~250 K).
Who first formulated entropy and the second law?
Sadi Carnot (1824, 'Reflections on the Motive Power of Fire') established that maximum heat-engine efficiency depends only on the reservoir temperatures, foreshadowing the second law without using entropy. Lord Kelvin (William Thomson, 1851) gave the impossibility statement of complete heat-to-work conversion. Rudolf Clausius (1854, 1865) coined 'entropy' from Greek for 'transformation' and stated the universe's entropy tends to a maximum. James Clerk Maxwell (1867) introduced the statistical view via his velocity distribution. Ludwig Boltzmann (1872) proved the H-theorem and engraved S = k log W (Planck's notation, 1900) on his tombstone. Willard Gibbs (1878) generalized to ensembles. The second law thus emerged from a 50-year collective effort, 1824-1877.