Statistical vs Thermodynamic Entropy

Melt an ice cube and its entropy rises. Thermodynamics tells you how much, from heat and temperature. It never tells you why the liquid is more disordered at the molecular level. Ludwig Boltzmann answered that question. He connected the macroscopic entropy of classical thermodynamics to the number of microscopic arrangements consistent with a given macroscopic state. This bridge, carved on his tombstone in Vienna as $S = k \ln W$ , is one of the most consequential equations in science. This article builds the Boltzmann relation from the ground up. You will see what $W$ actually counts. Then you will connect it to the Clausius definition of entropy. Finally you will apply it to real numbers, including residual entropy in crystals, and see where the analogy to "disorder" breaks down.

Key takeaways

Boltzmann entropy: $S = k \ln W$ , where $W$ is the number of microstates consistent with a given macrostate.
Thermodynamic (Clausius) entropy is defined from heat and temperature: $dS = dq_{rev}/T$ .
The two definitions agree numerically and conceptually; statistical mechanics explains why the thermodynamic laws hold.
Larger $W$ means more ways to arrange energy and particles, so entropy measures multiplicity, not literal "messiness."
Boltzmann's constant $k = 1.381 \times 10^{-23}\ \text{J K}^{-1}$ converts a dimensionless count into joules per kelvin.
Key idea: entropy is a measure of the number of equivalent microscopic ways a system can realize one observable macroscopic state.

The core insight

Entropy is not a substance or a measure of chaos; it is $k \ln W$ , a precise count (via Boltzmann's constant) of how many indistinguishable microscopic configurations produce the same macroscopic description, so that systems evolve toward the macrostate that can be realized in overwhelmingly more ways.

The physical picture

Consider a macroscopic sample: a mole of gas, a crystal, a beaker of solution. Thermodynamics describes it with a handful of variables — $T$ , $P$ , $V$ , $n$ — the macrostate. But at any instant the system also has a precise microscopic configuration: which atom has which position and momentum, which spin points which way. This is the microstate. Enormously many microstates correspond to the same macrostate, because swapping the labels on two identical, indistinguishable gas molecules with the same speeds changes nothing observable. Boltzmann's insight was that the macroscopic property we call entropy is proportional to the logarithm of $W$ , the number of microstates compatible with a given macrostate. A gas confined to half a box has far fewer accessible microstates than the same gas free to roam the whole box — after all, every molecule's position is restricted. When the partition is removed, the number of accessible microstates increases enormously, so $W$ increases, and so does $S$ . Nothing has to be "pushed" toward disorder; the system simply spends essentially all its time in whichever macrostate has overwhelmingly the largest $W$ , because that macrostate is so much more probable. The word "disorder" is only a loose metaphor. What is really being counted is multiplicity: the number of distinguishable ways of distributing energy quanta, positions, or orientations among the particles while keeping the total energy, volume, and particle number fixed. A perfect crystal at absolute zero, with every atom in one unique ground-state arrangement, has $W = 1$ and therefore $S = 0$ — this is the microscopic content of the Third Law, developed fully in the next article.

Figure

Same macrostate, many microstates: gas expanding into vacuum

Figure

Distributing 4 quanta of energy among 3 oscillators: counting W

The core relations

S = k \ln W

where $k$ is Boltzmann's constant (1.381 × 10⁻²³ J K⁻¹) and $W$ is the number of microstates for the given macrostate.

dS = \frac{dq_{rev}}{T}

where $dq_{rev}$ is heat absorbed reversibly at temperature $T$ ; this is the Clausius (thermodynamic) definition.

\Delta S_{mix} = -nR \sum_i x_i \ln x_i

where $x_i$ are mole fractions of ideal gases mixed at constant $T,P$ ; this is the entropy of mixing recovered from statistical counting.

S_m(0) = R \ln W_0

where $W_0$ is the number of energetically equivalent ground-state arrangements per mole, giving residual entropy when $W_0 > 1$ .

Derivation AFrom microstate counting to the entropy of mixing

Goal. Show that the statistical definition $S = k \ln W$ reproduces the known thermodynamic result for mixing two ideal gases, confirming the two definitions describe the same physical quantity.

Step 1. Set up the counting problem. Take $N_A$ molecules of gas A and $N_B$ molecules of gas B, each initially confined to its own container of volume $V$ , then combined into a shared volume $2V$ at the same temperature and pressure. Before mixing, each gas occupies its own volume; after mixing, both gases share the full volume $2V$ .

Step 2. Relate volume accessible to number of microstates. For an ideal gas, the number of positional microstates available to $N$ molecules in volume $V$ scales as $W \propto V^{N}$ (each molecule's position can be assigned independently across the volume, to good approximation at low density). Expanding one component's accessible volume from $V$ to $2V$ therefore multiplies its $W$ by $2^{N}$ .

Step 3. Write the entropy change for each component. Using $S = k \ln W$ ,

\Delta S_A = k \ln\!\left(\frac{W_{2V}}{W_V}\right) = k \ln\!\left(\frac{(2V)^{N_A}}{V^{N_A}}\right) = N_A k \ln 2

and similarly $\Delta S_B = N_B k \ln 2$ for gas B expanding into the shared volume.

Step 4. Generalize to arbitrary mole fractions. If instead of a 1:1 split the final mixture has mole fraction $x_A = N_A/N$ and $x_B = N_B/N$ with $N = N_A + N_B$ , each gas expands from its own partial volume $x_i V_{total}$ into the full $V_{total}$ , giving a factor $1/x_i$ rather than $2$ :

\Delta S_i = N_i k \ln\!\left(\frac{1}{x_i}\right) = -N_i k \ln x_i

Step 5. Sum over components and convert to molar quantities. Summing over all species and using $N_i k = n_i R$ (since $N_A k = R$ per mole, with $N_A$ Avogadro's number):

\Delta S_{mix} = -\sum_i n_i R \ln x_i = -nR\sum_i x_i \ln x_i

This is exactly the thermodynamic entropy of mixing derived earlier from the ideal-gas free energy — obtained here purely by counting microscopic spatial arrangements.

\boxed{\Delta S_{mix} = -nR \sum_i x_i \ln x_i \ \text{, identical whether derived from } dq_{rev}/T \text{ or from } k\ln W}

Worked examples

Example 1 — Counting microstates for a tiny spin system. Consider 4 non-interacting magnetic dipoles, each either "up" or "down," with no external field (all arrangements have equal energy, so all are equally probable microstates of one macrostate: total system, unconstrained). Total number of microstates:

W = 2^4 = 16

Boltzmann entropy of this macrostate:

S = k \ln W = (1.381\times10^{-23}\ \text{J K}^{-1})(\ln 16)

\ln 16 = 2.773

S = 1.381\times10^{-23} \times 2.773 = 3.830\times10^{-23}\ \text{J K}^{-1}

Now restrict to the macrostate "exactly 2 up, 2 down." The number of ways to choose which 2 of the 4 spins are up is

W' = \binom{4}{2} = 6

S' = k \ln 6 = (1.381\times10^{-23})(1.792) = 2.475\times10^{-23}\ \text{J K}^{-1}

Sanity check: $W' = 6 < W = 16$ , so $S' < S$ , as expected — restricting to a specific balanced macrostate excludes some microstates (like all-up or all-down) and lowers the count, even though "2 up, 2 down" is actually the single most probable outcome among all possible up/down counts for 4 spins. This distinction — most probable individual outcome versus total count over all outcomes — is exactly why real systems (with $N \sim 10^{23}$ spins) show such sharply peaked, reproducible macroscopic behavior: the most probable distribution utterly dominates the sum over all distributions.

Example 2 — Residual entropy of carbon monoxide, comparing statistical and calorimetric values. Solid CO has a small third-law "residual entropy" because the nearly symmetric CO molecule can sit in a crystal site as CO or OC with almost equal energy, so the crystal freezes in with essentially random orientation rather than achieving a unique order at absolute zero.

Statistical prediction. If each of the $N_A$ molecules in one mole has 2 equally likely orientations, independent of its neighbors,

W_0 = 2^{N_A}

S_m(0) = R \ln 2 = (8.314\ \text{J K}^{-1}\text{mol}^{-1})(0.6931) = 5.76\ \text{J K}^{-1}\text{mol}^{-1}

Experimental comparison. The calorimetric (third-law) entropy of CO obtained by integrating $C_p/T$ from near 0 K up to the gas phase, and comparing with the entropy computed independently from spectroscopic/statistical thermodynamic data (which assumes perfect order at $T=0$ ), shows a discrepancy of about

S_m(0)_{exp} \approx 4.6\ \text{J K}^{-1}\text{mol}^{-1}

Sanity check: the observed residual entropy (≈4.6 J K⁻¹ mol⁻¹) is smaller than the naive $R\ln 2 = 5.76$ J K⁻¹ mol⁻¹ predicted for fully random, independent orientation, because real CO crystals have some residual short-range correlation between neighboring dipoles that slightly reduces the true count of accessible microstates below $2^{N_A}$ . The order of magnitude and the very existence of a nonzero residual entropy, however, is exactly what the simple two-orientation counting model predicts — a clean, checkable link between $S = k\ln W$ and calorimetry.

Limiting cases & common traps

Situation	What happens	Why
Perfect crystal at $T \to 0$ K	$W \to 1$ , so $S \to 0$	Only one unique ground-state arrangement is accessible
Gas expanding into vacuum	$W$ increases sharply	Vastly more positional microstates become available
Frozen-in molecular disorder (CO, $\text{H}_2\text{O}$ ice)	$S(0) > 0$	Orientational choices get "locked in" before equilibrating, so $W_0 > 1$
Very large $N$	Most probable macrostate totally dominates	Multiplicities peak so sharply that fluctuations are unobservable
Comparing $S$ for two states	Only $\ln(W_2/W_1)$ matters	Absolute $W$ is uncomputable for macroscopic systems; ratios are what count

The chemist behind it

Ludwig Boltzmann spent much of his career defending the atomic hypothesis against skeptical contemporaries (notably Ernst Mach and the energeticists) who doubted that atoms were real, physical entities rather than convenient bookkeeping devices. His statistical interpretation of entropy, developed through the 1870s–1890s, required accepting that gases really are collections of discrete, countable particles whose arrangements could, in principle, be enumerated. The formula $S = k \ln W$ (the constant $k$ was named in his honor by Max Planck, who first wrote the relation in this precise form) turned the abstract, purely macroscopic entropy of Clausius into something with a concrete, calculable microscopic meaning. Boltzmann did not live to see atomic theory fully vindicated by Einstein's and Perrin's work on Brownian motion; he died in 1906, and the equation is engraved on his gravestone in Vienna's Zentralfriedhof.

Origins · 1 minLudwig Boltzmann and how this was discovered

Ludwig Boltzmann

1844–1906 · linked microscopic particle arrangements to macroscopic entropy via S = k ln W

Ask the tutor: How does the Third Law reconcile perfect crystals with real substances that keep a nonzero residual entropy down to absolute zero?

Is W literally a number you can compute for a real sample of gas

In principle yes, in practice no. $W$ for a mole of gas is astronomically large — of order $10^{10^{23}}$ or larger — far beyond any numerical computation or even meaningful comparison to familiar numbers. What matters physically is never the absolute value of $W$ , but the ratio $W_2/W_1$ between two states, since $\Delta S = k\ln(W_2/W_1)$ is what's measurable and what appears in every thermodynamic calculation. Statistical mechanics gives formulas (partition functions) for computing these ratios without ever evaluating $W$ itself.

Does S = k ln W apply only to isolated systems

The simplest form, with $W$ counted at fixed total energy, applies to an isolated system (the microcanonical ensemble). For systems in thermal contact with a bath at temperature $T$ (the canonical ensemble), entropy is instead computed from the partition function $q$ , via $S = U/T + k\ln q$ for a single particle or the corresponding molar expression. These are equivalent, consistent formulations; the microcanonical picture is simply the most transparent way to first introduce the logic of $S=k\ln W$ .

How does this connect to the Third Law of Thermodynamics

The Third Law states that the entropy of a perfect crystal approaches zero as $T \to 0$ K. Statistically this is because a perfect, defect-free crystal in its unique lowest-energy configuration has exactly one accessible microstate, so $W=1$ and $S = k\ln 1 = 0$ . Real crystals sometimes retain small residual entropies (like CO or ice) because some disorder gets kinetically trapped before the system can fully order — this is explored in depth in the dedicated Third Law article.

Why is entropy of mixing always positive for distinguishable gases

Because combining two initially separated gases strictly increases the number of accessible positional microstates: molecules of each species now have access to the entire combined volume rather than just their original share. Since $\ln x_i < 0$ for any mole fraction $x_i < 1$ , every term $-nR\,x_i\ln x_i$ is positive, so $\Delta S_{mix} > 0$ for any nontrivial mixture. This holds regardless of whether the gases interact, as long as they are genuinely distinguishable species.