Home > Positioning > Subjects > Thermodynamics > The laws and the reduction

The Laws and the Reduction

The four laws

The zeroth law. If two systems are each in thermal equilibrium with a third, they are in equilibrium with each other. It was numbered last — Ralph Fowler in the 1930s, after the first three had their numbers — once it was recognised that this transitivity is what makes temperature definable at all, and therefore logically prior to the laws that use the concept.

The first law. Energy is conserved; heat and work are two ways of transferring it. This is the law that makes energy accounting possible and rules out perpetual motion of the first kind.

The second law. The one that carries the field’s weight and the one that causes its arguments. In Clausius’s version heat does not pass spontaneously from a colder body to a hotter one; in Kelvin’s, no cyclic process converts heat entirely into work. Entropy — Clausius’s coinage, 1865 — never decreases in an isolated system.

The third law. Absolute zero cannot be reached in a finite number of steps; entropy approaches a constant as temperature approaches zero.

That there are four laws is less settled than the numbering suggests. Uffink’s survey finds the Clausius, Kelvin, Planck, Carathéodory and Lieb–Yngvason formulations of the second law are not straightforwardly equivalent to one another, so “the second law” names a family whose members do different work.

The statistical reading

Boltzmann and Gibbs built the account intended to underwrite the laws from below. Boltzmann’s entropy is the logarithm of the number of microscopic states compatible with a macroscopic description — entropy as a count. Gibbs’s is defined over a probability distribution across microstates, and is the version that generalises.

The two are not interchangeable, and the difference became consequential when small systems did. They agree in the thermodynamic limit and diverge for systems of few particles. Sheldon Goldstein, Joel Lebowitz, Roderich Tumulka and Nino Zanghì argued in 2020 that they agree to leading order for macroscopic systems in equilibrium; Charlotte Werndl and Roman Frigg dispute that reconciliation. A separate exchange runs over which is correct for small systems: Jörn Dunkel and Stefan Hilbert argued in 2014 that the Gibbs volume entropy is the right one and that negative absolute temperatures are an artefact of using Boltzmann’s; Daan Frenkel and Patrick Warren replied that the Gibbs entropy fails the zeroth-law requirement that bodies in equilibrium share a temperature, which Boltzmann’s satisfies.

One technical fact drives much of what follows: the Gibbs entropy is a constant of the motion. Under the underlying dynamics it cannot increase — this is a theorem, not an interpretive position. The standard response is coarse-graining, but coarse-grained entropy increases only if the system is mixing, a condition Uffink describes as very demanding and which many real systems fail.

Irreversibility, and what the H-theorem does not prove

Boltzmann’s H-theorem appears to derive the approach to equilibrium from mechanics. Two objections from his own century established that it does not do so alone.

Loschmidt’s reversibility objection (1876): the underlying mechanics is time-symmetric, so for every trajectory in which entropy rises there is a reversed trajectory in which it falls. No time-asymmetric conclusion can follow from time-symmetric premises.

Zermelo’s recurrence objection (1896), from Poincaré recurrence: a bounded mechanical system returns arbitrarily close to any earlier state given enough time, so entropy cannot increase monotonically forever.

What the theorem actually uses is the Stosszahlansatz, the molecular chaos assumption — that the velocities of colliding particles are uncorrelated before collision but not after. That assumption is itself time-asymmetric, so the asymmetry is put in rather than derived. The modern reading is that the H-theorem establishes a statistical tendency, not a mechanical necessity, and Boltzmann’s own position moved toward the combinatorial account plus an appeal to cosmological initial conditions.

Where the asymmetry then comes from is open. The past hypothesis — that the universe began in an extremely low-entropy state, developed by David Albert and Barry Loewer — is the best-known answer but is one position among several rather than a settled result. John Earman’s “The ‘Past Hypothesis’: Not even false” (2006) argues it cannot be given a precise, coordinate-independent formulation in general relativity. Huw Price presses a different charge, that the standard framing presupposes the asymmetry it sets out to explain. Typicality approaches, dynamical-collapse approaches, and the perspectival reading described on the entropy page are the other live options.

Does thermodynamics reduce to statistical mechanics?

“Thermodynamics reduces to statistical mechanics” is a slogan covering at least three separable claims, and much of the dispute turns on which is meant:

Every thermodynamic quantity has a statistical-mechanical counterpart — temperature and mean kinetic energy, say.
The laws of thermodynamics are derivable from statistical mechanics, the second law in particular.
Thermodynamics retains no autonomous explanatory value once (1) and (2) are done.

One can hold (1) and deny (3), and several people do.

On the reductionist side, Steven Weinberg’s Dreams of a Final Theory (1992) takes the strong line that higher-level science follows as a corollary once the fundamental level is settled. Albert’s Time and Chance (2000), and the Mentaculus programme he developed with Loewer, treat statistical mechanics as foundational across the sciences.

Craig Callender’s “Taking Thermodynamics Too Seriously” (2001) is easy to misread in either direction. His target is not statistical mechanics but the habit of treating thermodynamic laws as literal truths requiring reduction: he argues philosophers should give up part of the reduction project and deny that thermodynamics is “universally true and somehow independent of the statistics of the micro-constituents.” The position is deflationary about thermodynamics rather than triumphant about reduction.

On the anti-reduction side, Robert Batterman’s The Devil in the Details (2002) argues that phase transitions resist reduction because their explanation runs through a singular limit: asymptotic explanation works by deleting information rather than supplying more of it, which is a different thing from micro-explanation. John Norton answers directly that no actual infinity is required, leaving the reduction route open. The exchange is unresolved.

A middle route runs through functionalism. Katie Robertson proposes defining thermodynamic quantities by their nomological role and then asking what plays that role in statistical mechanics — if mean kinetic energy plays the temperature role, temperature functionally reduces to it. The claimed advantage is that functionalism specifies which mismatches between the two theories can be tolerated. David Wallace’s “Asymmetry, Abstraction, and Autonomy” (2020) argues on technical grounds that coarse-grained dynamics can be genuinely autonomous rather than merely convenient. David Lavis, Reimer Kühn and Frigg argue that even a successful reduction would not exhaust what thermodynamics does.

Uffink and Frigg are best read as cartographers of this dispute rather than parties to it; Frigg’s “A Field Guide to Recent Work on the Foundations of Statistical Mechanics” (2007) is the standard survey.