Drop a single drop of ink into a glass of still water and walk away. Come back an hour later and the ink has spread evenly through the whole glass — a uniform pale grey. You have never seen the reverse: a grey glass spontaneously gathering its ink back into one sharp drop. The molecules would be perfectly happy to do it (the laws governing each collision run equally well forwards and backwards), and no law of force forbids it. Yet it never happens. Why?

The answer is not a new force. It is counting. There are almost unimaginably more ways for the ink to be spread out than to be gathered up, so a randomly wandering system spends essentially all of its time spread out. That single idea — that the macroscopic behaviour of matter is decided by how many microscopic arrangements look the same to us — is the engine of the whole of statistical mechanics. This page builds it from scratch: the distinction between a macrostate and its many microstates, the count \Omega of those microstates, and the one short equation that turns a headcount into a thermodynamic quantity — Boltzmann's entropy, S = k_B \ln \Omega.

Four coins: the whole idea in miniature

Forget 10^{23} molecules for a moment and flip four coins. Suppose the only thing you bother to record is how many came up heads — call that number the macrostate. You do not write down which coins; you only care about the total. The full record of every individual coin (H, T, T, H, …) is a microstate.

There are 2^4 = 16 equally likely microstates in all. Group them by macrostate (number of heads) and count:

The counts 1,4,6,4,1 are exactly the binomial coefficients \binom{4}{k}, and they add to 16. Now read the table as physics. Each microstate is equally likely (that is the only assumption we make), so the probability of a macrostate is just its share of the microstates. The "two heads" macrostate has 6/16 of them; "all heads" has a lonely 1/16. The balanced, boring, middle macrostate wins simply because there are more ways to build it. Blow the four coins up to 10^{23} gas molecules choosing the left or right half of a box, and "lonely" becomes "never in the lifetime of the universe".

Macrostate, microstate, multiplicity

Let us pin the three words down, because getting them straight is nine-tenths of the subject.

And one assumption ties them together — the load-bearing postulate of the whole theory:

Notice what this does not say. It does not say the system prefers equilibrium, or is pulled toward it, or minimises anything. It says all microstates are on an equal footing, and equilibrium wins by sheer weight of numbers. Everything else — the arrow of time, the second law, why heat flows hot-to-cold — is downstream of that one democratic assumption.

Counting a gas: particles in two halves of a box

Replace the coins with a gas. Imagine a box split, in our imagination only, into a left half and a right half. Each of the N molecules is, at any instant, in one half or the other — that is our stand-in for "heads or tails". The macrostate is how many molecules, n, are in the left half. The number of microstates realising it is the number of ways to choose which n of the N molecules sit on the left — the binomial coefficient:

Worked count, N = 4. The all-left macrostate n=4 has \Omega = \binom{4}{4} = 1 — there is exactly one way for every molecule to be on the left. The even split n=2 has \Omega = \binom{4}{2} = \tfrac{4!}{2!\,2!} = 6. So the balanced macrostate is already 6 times more likely than the completely-lopsided one, with only four molecules.

Now let N get large. The even split n = N/2 always carries the biggest \Omega, and its lead grows explosively. For N = 100 the ratio of the even-split multiplicity to the all-on-one-side multiplicity is \binom{100}{50} \approx 10^{29} to 1. For a real, mole-sized N \approx 6\times10^{23} the odds against finding even a one-part-in-a-billion excess on one side are so steep that you would wait many times the age of the universe to see it once. This is why a gas fills its container uniformly: not because it is forced to, but because "uniform" is where almost all of the microstates live.

The distribution of \Omega(n) across macrostates is a binomial peak. Slide N up in the graph below (it plots the multiplicity relative to its maximum, \Omega(n)/\Omega_{\max}, against the fraction n/N) and watch the bump around the even split grow into a needle-sharp spike.

The width of that peak scales like \sqrt{N}, but the position scale is N itself, so the relative fluctuation of the split away from 50–50 shrinks as

For N = 10^{23} that relative wobble is about 10^{-11.5} — utterly unmeasurable. The macroscopic world looks sharp and deterministic precisely because N is so absurdly large.

From a headcount to entropy: Boltzmann's formula

Multiplicities are monstrous numbers — a mole of gas has an \Omega with something like 10^{24} digits. Worse, they behave badly when you combine systems. Put system A (with \Omega_A microstates) beside an independent system B (with \Omega_B). Each microstate of A can pair with each of B, so the combined multiplicity multiplies:

But every macroscopic property we know — energy, volume, mass — adds when you put two systems together; such quantities are called extensive. We would love our new state-variable to add too. There is exactly one function that turns multiplication into addition — the logarithm. That is the whole reason for the \ln in Boltzmann's definition of entropy:

The equation S = k_B \ln \Omega is carved on Boltzmann's gravestone in Vienna (in the form S = k \log W). It is arguably the single most important sentence in all of thermal physics: it connects the microscopic world of countable arrangements to the macroscopic world of thermometers and heat engines.

A concrete number. Take a tiny system of N = 100 two-state particles, all microstates accessible, so \Omega = 2^{100}. Then

Tiny — because 100 particles is tiny. Scale N up to Avogadro's number and S climbs to the everyday joules-per-kelvin you meet in a thermodynamics table. Note the clean pattern: for N two-state particles, S = N k_B \ln 2 — the entropy is simply \ln 2 per particle (one "bit" each), times k_B.

The second law, as a statement about counting

Now the payoff. Release a gas that starts crammed into the left half of a box (a low-multiplicity, and therefore low-entropy, macrostate). Molecules bounce around at random. At every instant the system is equally likely to be in any accessible microstate — and there are vastly more microstates with the gas spread through the whole box than clustered on one side. So overwhelmingly the system is found spread out, and it stays spread out. The macrostate drifts toward larger \Omega, hence larger S — not because anything pushes it, but because that is where almost all the possibilities are.

How improbable? For a gas of N molecules to fluctuate all the way back into one half of the box, the odds are 1 in 2^{N}. For a mere N = 100 that is already 1 in 10^{30}; for a real gas it is 1 in 2^{(10^{23})}, a number so large that writing its digits would fill more paper than there are atoms in the galaxy. The system could reverse — the Poincaré recurrence theorem even guarantees it eventually returns arbitrarily close to its start — but the expected waiting time dwarfs the age of the universe by factors that make the word "astronomical" look shy. The second law is not a certainty carved into the microphysics; it is a near-certainty carved into the arithmetic of large numbers.