THERMODYNAMICS · STORY 05

Entropy is not
disorder

It is an honest count. The moment the valve opens, the numbers take over.

HOW TO READ

The stage on the right is driven by one set of particles: the macroscopic bulbs show what the gas does, the histogram counts how many ways each split can happen, and S = k ln W turns the count into thermodynamics. The three panels never disagree — entropy is not a quantity from another world; it is this count.

One census · three representations

N: 100
n: 100
ln W: 0.0

Macro · Bulbs Particle positions are schematic; the left-right counts match every statistic exactly.

ln W

S / k = ln W = 0.0

Odds of full return ≈ 10^-30

Symbol · S = k ln W ln W in natural log; entropy reported in units of k.

W(n) is computed point by point from binomial coefficients; the axis shows probability.

Two glass bulbs, a valve between them. Gas on the left, vacuum on the right.

Open the valve — the gas roars into the right bulb, and within seconds each side holds half. Wait a day, a year, a hundred million years: it will never crawl back into the left bulb on its own.

No law of mechanics forbids the return. Play every collision backwards and each one is perfectly legal. So what is standing in the way?

Place your bet.

Predict first: what stops the gas from returning?

Forget "messiness" — let us count.

Give every molecule a coin: heads = left bulb, tails = right. Each heads-tails pattern of the $N$ molecules is one microstate, and the number of ways to put exactly $n$ molecules on the left is

W(n) = \binom{N}{n} = \frac{N!}{n!\,(N-n)!}

With $N=4$ : all-left can happen in exactly 1 way; a two-two split in 6. Drag the particle count and watch the census take off — by $N=100$ , the even split has about $10^{29}$ ways, while "all left" still has its lonely 1.

Particles N N = 4

Plot $W(n)$ and you do not get a gentle hill. You get a spike.

The bigger $N$ , the sharper: the peak's width grows only as $\sqrt{N}$ while the axis stretches as $N$ . At $N=100$ , a window of ±10 molecules already holds over 95 % of all microstates; at $N=10^{23}$ the window holding 99.9999…% is too narrow to draw — thinner than an atom.

The macroscopic world only ever sees the tip of the spike. "Equilibrium" is that spike's name.

Boltzmann carved this census into a formula — and onto his gravestone:

S = k \ln W

The logarithm is there because $W$ multiplies for independent systems while thermodynamics needs entropy to add; the constant $k = 1.38\times 10^{-23}\ \mathrm{J/K}$ converts a count into calorimeter units.

Double the volume of one mole of gas and every molecule doubles its options, so $W$ gains a factor $2^{N_A}$ :

\Delta S = N_A k \ln 2 = R\ln 2 = 5.76\ \mathrm{J\,K^{-1}\,mol^{-1}}

Not a metaphor — a number a calorimeter can check.

Now answer the opening question: why does the gas never come back?

Press the button: all 100 molecules are stuffed back into the left bulb, then released. Each step, one randomly chosen molecule hops sides (Ehrenfest's urn). Watch the count fall toward the spike — and get pinned there, jittering.

Returning to all-left is not forbidden; its probability is $2^{-100} \approx 10^{-30}$ . Try once per nanosecond and you would wait $10^4$ ages of the universe. For a mole of molecules the exponent becomes $-10^{23}$ .

Irreversibility is not a decree. It is odds.

In small systems, though, the odds are not lopsided — watch the histogram: the count keeps jittering.

At $N=100$ , ±10 % excursions happen all the time; Brownian motion and critical opalescence are entropy's fluctuations made visible. Their relative size shrinks as $\sqrt{N}/N = 1/\sqrt{N}$ : ten-thousand-fold more particles, hundred-fold quieter.

The iron certainty of the "laws" of thermodynamics is a gift statistics hands over as $N\to 10^{23}$ .

Finally, settle the famous metaphor: entropy is not "mess".

Oil and water separating into neat layers is the higher-entropy state — in the hydrophobic effect it is the water molecules' count of arrangements that rules
Hard spheres at high density crystallize spontaneously: the tidy lattice gives every sphere more room to rattle, $W_{\text{crystal}} > W_{\text{jumbled}}$
A "messy-looking" deck of cards and a suit-sorted one have exactly the same number of microstates — messiness is human aesthetics; entropy only counts

"Disorder" has guessed right many times, and every time it guesses wrong it is because it was never the definition. There is only one definition: $S = k\ln W$ .

The counter is yours.

Drag the particle count and watch the spike sharpen; frame a window and see how many microstates it swallows; herd every molecule back to the left and watch the odds drive them home to the spike.

Challenge: find an $N$ where a ±5 % window captures exactly 99 % of all microstates — then imagine pushing it to $10^{23}$ .

Particles N N = 100 Window half-width ± ±10 A ±10 window captures 96.5% of all microstates

THE LEDGER

The counter’s ledger

Microstates W The number of microscopic ways to realize one macroscopic look; in the two bulbs W(n) = C(N, n), countable one by one.

Boltzmann’s formula S = k ln W: the log makes entropies of independent systems add, and k converts a count into calorimeter units. Carved on his grave in Vienna.

The √N law Fluctuations grow as √N in absolute size and vanish as 1/√N in relative size — thermodynamic certainty is a gift of large numbers.

R ln 2 The 5.76 J/(K·mol) entropy of doubling a mole of gas: derived by counting, checkable by calorimetry.

CONCEPT CHECK · PREDICT, THEN TEST

At constant temperature, one mole of ideal gas is compressed back to half its volume. What happens to the system’s entropy?

Nothing — temperature did not change, and entropy depends only on temperature It drops by R ln 2 — each molecule’s options halve, but the surroundings gain at least as much It rises — the compression work put energy into the gas It drops, and the entropy of the whole universe drops with it

FREE PLAY · Leave the storyline with a question

The counting laboratory

Particle count, observation window, release experiment — all yours; the three panels answer together.

Particles NN = 100 Window half-width ±±10

A ±10 window captures 96.5% of all microstates

One census · three representations

N: 100
n: 50
ln W: 66.8

Macro · Bulbs Particle positions are schematic; the left-right counts match every statistic exactly.

ln W

S / k = ln W = 66.8

Odds of full return ≈ 10^-30

Symbol · S = k ln W ln W in natural log; entropy reported in units of k.

W(n) is computed point by point from binomial coefficients; the axis shows probability.

MODEL CARD

Model card · V0.1

System

A two-equal-bulb lattice gas: N in-principle-distinguishable particles, each equally likely left or right, W(n) = C(N, n). The minimal microstate-counting model of the classroom; real gases add momentum dimensions to the count, making the verdicts even more lopsided.

Computation

All statistics run in log space (Stirling with the 1/(12n) correction for n ≥ 32, relative error < 1e-10), so N reaches hundreds without overflow; the histogram normalizes point by point, and unit tests pin normalization, symmetry, the 2⁻ᴺ tail, and R ln 2 = 5.763 J/(K·mol).

Release animation

Ehrenfest’s urn: each step one particle, chosen by a fixed-seed LCG, hops sides — reproducible by construction. A cartoon of relaxation, not molecular dynamics; real relaxation rates are set by collision frequencies.

Sources for the "tidy yet higher-entropy" cases

The hydrophobic effect (water’s configuration count drives demixing) and hard-sphere crystallization (the Alder transition, first shown by computer experiment in 1957) are standard statistical mechanics; this page cites them qualitatively — numerical simulations are backend work, see docs/后端待办.md.

What cannot be inferred from this page

Nothing here treats the entropy of energy exchange (temperature, heat capacities, the third law) or the relation between information entropy and thermodynamic entropy — other stories. Mixing identical gases needs the Gibbs-paradox correction to particle distinguishability, declared here honestly.

Entropy is not a substance, not a force, and not "mess". It is the number of ways — and large numbers show no mercy.

Next stop: let this census drive a chemical reaction downhill.

Back to the story catalog GitHub ↗

Entropy is notdisorder