Microstates and Macrostates

Pull a single molecule out of the air in front of you and ask where it is and how fast it is moving. Now ask it of all $2.5\times10^{25}$ molecules in this cubic metre at once. That complete list — every position, every velocity — is a microstate, and it is hopeless: it changes completely a trillion times before you finish writing the first number. Yet the air has a temperature, a pressure, a density that sit rock-steady while that microscopic storm rages underneath. How can something so stable ride on something so chaotic? The answer is the founding move of statistical mechanics: split the description in two. The microstate is the full microscopic list; the macrostate is the short one — energy, volume, particle number, maybe pressure and temperature — that you can actually measure. The crucial fact is that a single macrostate is compatible with an astronomical number of microstates. Count them, call the count $\Omega$ , and you hold the keystone of the whole subject: its logarithm is the entropy, $S=k_{\mathrm B}\ln\Omega$ , and from there all of thermodynamics unfolds. Why does averaging over microstates give sharp, repeatable numbers instead of noise? Because the counts are so wildly unequal. One macrostate — the equilibrium one — is realized by so overwhelmingly many more microstates than any competitor that the system is found in it with effective certainty, and fluctuations away shrink as $1/\sqrt{N}$ . Master this microstate–macrostate bookkeeping and you own the logic behind entropy, irreversibility, the Gibbs paradox, and every ensemble you will meet: they are all ways of counting microstates.

Key takeaways

A microstate is the complete instantaneous specification of every dynamical variable — for $N$ classical particles, the $6N$ numbers $(q,p)$ , i.e. one point in phase space.
A macrostate is fixed by only a handful of macroscopic observables ( $E,V,N$ , or $T,P,\mu$ ) and is realized by an enormous number $\Omega$ of microstates.
The multiplicity $\Omega$ is astronomically large; its logarithm is the Boltzmann entropy $S=k_{\mathrm B}\ln\Omega$ , the bridge from counting to thermodynamics.
Statistical mechanics works because one macrostate dominates the count so completely that relative fluctuations scale as $1/\sqrt{N}\sim10^{-12}$ for a mole.
Irreversibility is microstate counting in disguise: systems flow from macrostates of small $\Omega$ to macrostates of overwhelmingly larger $\Omega$ .
Key idea: a macrostate is an equivalence class of microstates; statistical mechanics computes the size $\Omega$ of that class, and $S=k_{\mathrm B}\ln\Omega$ turns the count into thermodynamics.

The core insight

A macrostate is an equivalence class of microstates — many microscopic configurations that look identical through a few macroscopic observables. The number $\Omega$ of microstates in that class is what statistical mechanics computes, and $S=k_{\mathrm B}\ln\Omega$ converts the count into thermodynamics.

Where this lives today

The microstate–macrostate split is not a historical relic; it is the operating principle of modern simulation and of the physics of information. A molecular-dynamics run tracks one microstate trajectory through $6N$ -dimensional phase space and reports macrostate averages — temperature, pressure, diffusion constants — exactly as this article prescribes, and its correctness rests on the assumption that the sampled microstates fairly represent the macrostate's $\Omega$ . The deepest modern echo is Landauer's principle. Erasing one bit of information forces two microstates into a single macrostate, and that irreversible loss of counting must dump at least $k_{\mathrm B}T\ln2$ of heat — a bound now measured directly in single-electron and colloidal experiments and taken seriously by the designers of low-power and reversible computers. The shared functional $-\sum p_i\ln p_i$ that appears in Gibbs's entropy and Shannon's information is no coincidence: both are counting the same thing.

The physical picture

Shuffle a deck of cards. A microstate is the exact order of all 52; a macrostate is a coarse property like "well shuffled." There is exactly one perfectly sorted order but astronomically many disordered ones, so shuffling reliably destroys order — not because any single ordering is favoured, but because the disordered macrostate simply owns far more microstates. Equilibrium is nothing more mysterious than the macrostate with the most microstates. The same accounting runs a gas. Divide a box into a left and a right half and ask how many molecules sit on the left. "All $N$ on the left" is one macrostate; "half on each side" is another. The uniform macrostate is realized by vastly more arrangements, so the gas fills the box. The figure below counts them for a small system where you can still see the spread, then shows what happens as $N$ grows: the distribution collapses onto a needle at the uniform value.

Figure

Counting microstates: how many ways to put n of N=40 particles in the left half

Read the figure as a competition for microstates. The uniform macrostate ( $n=N/2$ ) wins by a landslide even at $N=40$ ; the extreme macrostates ("all on one side") are single arrangements, invisible on this scale. Now watch the punchline as the system grows — the spread narrows relentlessly, because the relative width of the peak scales as $1/\sqrt{N}$ :

Figure

As N grows the equilibrium macrostate becomes a needle — fluctuations vanish as 1/√N

That collapse is the whole reason thermodynamics is sharp. The microscopic dynamics is reversible and deterministic — every trajectory has a valid time-reverse — yet the macroscopic arrow of time appears because systems drift from macrostates of few microstates to macrostates of many. The coarse-graining lives in our description, not in nature.

The core equations

A classical microstate is the full list of coordinates and momenta. For $N$ particles in three dimensions that is $6N$ numbers, assembled into a single point $\Gamma$ in phase space. This one object carries everything the microscopic laws need:

\Gamma=(q_1,\dots,q_{3N},\,p_1,\dots,p_{3N}) \qquad (1)

where: $\Gamma$ — a single point in $6N$ -dimensional phase space, i.e. one microstate; $q_i$ — a generalized coordinate; $p_i$ — its conjugate momentum; $N$ — particle number ( $3$ spatial degrees of freedom each). Nothing about the macroscopic world is missing from $\Gamma$ ; the problem is only that it has too many numbers to know. The microstate is not static — it moves. Hamilton's equations give its velocity in phase space, and because they are first-order in time, the present microstate fixes the entire past and future:

\dot q_i=\frac{\partial H}{\partial p_i},\qquad \dot p_i=-\frac{\partial H}{\partial q_i} \qquad (2)

where: $H(q,p)$ — the Hamiltonian, total energy as a function of the microstate. Determinism here has a sharp geometric consequence: two microstates can never share a future, so phase-space trajectories never cross. That single fact underlies Liouville's theorem and the incompressible flow of probability. A macrostate, by contrast, is a short list. For an isolated system it is just the conserved data:

\text{macrostate}=\{E,\,V,\,N\}\ \ (\text{or }T,P,\mu,\dots) \qquad (3)

where: $E$ — total energy; $V$ — volume; $N$ — particle number. These few labels are all you measure; every microstate consistent with them belongs to the same macrostate. The bridge between the two levels is the multiplicity $\Omega$ — the number of microstates a macrostate contains. Classically it is a phase-space volume made dimensionless and de-permuted:

\Omega(E,V,N)=\frac{1}{N!\,h^{3N}}\int \mathrm{d}^{3N}q\,\mathrm{d}^{3N}p\ \delta\!\big(E-H\big)\,\delta E \qquad (4)

where: $\Omega$ — count of accessible microstates; $h^{3N}$ — the volume of one quantum phase-space cell, so $\Omega$ is a pure number; $N!$ — the Gibbs factor, because permuting identical particles is not a new microstate; $\delta E$ — a thin shell width, $\delta E\ll E$ . Drop the $1/N!$ and entropy stops being extensive — the Gibbs paradox — so these factors are physics, not bookkeeping. For an isolated system the postulate of equal a priori probabilities assigns every accessible microstate the same weight:

P(\Gamma)=\frac{1}{\Omega}\quad\text{for every accessible }\Gamma \qquad (5)

where: $P(\Gamma)$ — probability of microstate $\Gamma$ . Uniform weight is the least-biased choice consistent with what we know — and, as the second derivation shows, it is exactly the distribution that maximizes entropy. The multiplicity connects to thermodynamics through Boltzmann's relation, the equation on his tombstone:

S(E,V,N)=k_{\mathrm B}\ln\Omega \qquad (6)

where: $S$ — thermodynamic entropy; $k_{\mathrm B}=1.380649\times10^{-23}\ \mathrm{J/K}$ (exact, 2019 SI). The logarithm is what makes entropy additive: independent systems multiply their microstate counts, $\Omega=\Omega_1\Omega_2$ , and the log turns the product into a sum, $S=S_1+S_2$ . Finally, why the single equilibrium macrostate is seen with certainty: fluctuations of extensive quantities are tiny for large $N$ ,

\frac{\sqrt{\langle (\Delta E)^2\rangle}}{\langle E\rangle}\sim \frac{1}{\sqrt{N}} \qquad (7)

where: $\Delta E$ — energy fluctuation about the mean. For a mole, $1/\sqrt N\sim10^{-12}$ : far below any macroscopic instrument, so the most probable macrostate is the observed one.

Derivation AWhy the equilibrium macrostate is a needle

Here we prove the claim the figures made visually: for a system split into two halves, the number of microstates with a given occupation is a binomial coefficient, its maximum sits at exactly uniform density, and the relative width of the peak shrinks as $1/\sqrt N$ . The strategy is to count microstates directly, find the peak by differentiating the logarithm, and expand around it with Stirling's approximation — turning "equilibrium is the most probable macrostate" into a sharp statement about a Gaussian whose width vanishes in the thermodynamic limit.

Step 1. Label the macrostate by $n$ , the number of particles in the left half. Each choice of which $n$ particles are on the left is a distinct microstate, so the multiplicity is $\Omega(n)=\binom{N}{n}$ . Why: the macrostate fixes only the count $n$ , not which particles; the number of ways to choose them is the binomial coefficient.

Step 2. The total over all macrostates is $\sum_{n=0}^{N}\binom{N}{n}=2^N$ (each particle independently left or right), so equal a priori probabilities give $P(n)=\binom{N}{n}/2^N$ . Why: every microstate carries equal weight $1/2^N$ , and $P(n)$ collects the $\Omega(n)$ microstates sharing that $n$ .

Step 3. Locate the peak by setting $\frac{d}{dn}\ln\binom{N}{n}=0$ . Using $\ln\binom{N}{n}\approx N\ln N - n\ln n-(N-n)\ln(N-n)$ (Stirling), the derivative is $\ln\!\frac{N-n}{n}=0$ , so $n^*=N/2$ . Why: the most probable macrostate is uniform density — the equilibrium state — and it falls straight out of maximizing the log-multiplicity.

Step 4. Expand to second order: $\ln\Omega(n)\approx\ln\Omega(n^*)-\frac{2}{N}(n-n^*)^2$ , a Gaussian with variance $\sigma^2=N/4$ . The relative width is

\frac{\sigma}{N}=\frac{1}{2\sqrt{N}} \qquad (8)

Why: the absolute spread grows only as $\sqrt N$ while the range grows as $N$ , so the fraction on the left is pinned to $1/2$ ever more tightly — for $N\sim10^{23}$ the relative width is $\sim10^{-12}$ , and no measurement can resolve a departure from uniformity. *Verified: binomial concentration; peak, variance, and the $1/\sqrt N$ law reproduced symbolically and numerically in verify_aq-sm-microstates-and-macrostates.py. Cf. Reif §1, Kardar Ch. 4.*

Go deeper · Derivation B · 1 minBoltzmann's entropy is Gibbs's, maximized

Derivation BBoltzmann's entropy is Gibbs's, maximized

This is the article's owned structural result: the general Gibbs entropy functional, maximized over all probability distributions on the $\Omega$ accessible microstates subject only to normalization, is stationary exactly at the uniform distribution — and there it equals $k_{\mathrm B}\ln\Omega$ . So the postulate of equal a priori probabilities is not an extra assumption bolted on; it is the least-biased (maximum-entropy) distribution, and Boltzmann's $S=k_{\mathrm B}\ln\Omega$ is the special case of Gibbs's $S=-k_{\mathrm B}\sum p_i\ln p_i$ on that distribution.

Step 1. Build the Lagrange functional $\mathcal L=-k_{\mathrm B}\sum_i p_i\ln p_i-\lambda(\sum_i p_i-1)$ , with $\lambda$ enforcing $\sum_i p_i=1$ . Why: we maximize the Gibbs entropy over the $\Omega$ microstates with the single constraint that the probabilities sum to one — the only information an isolated system gives us.

Step 2. Stationarity $\partial\mathcal L/\partial p_i=0$ gives $-k_{\mathrm B}(\ln p_i+1)-\lambda=0$ , so $p_i=\exp(-1-\lambda/k_{\mathrm B})$ — a constant independent of $i$ . Why: the derivative condition has the same value for every microstate, so the maximizing distribution must be flat.

Step 3. Normalization then forces $p_i=1/\Omega$ , and substituting back,

S=-k_{\mathrm B}\sum_{i=1}^{\Omega}\tfrac1\Omega\ln\tfrac1\Omega=k_{\mathrm B}\ln\Omega \qquad (9)

Why: the uniform distribution IS the equal-a-priori-probabilities postulate, here derived rather than assumed; evaluating the Gibbs functional on it collapses the sum and reproduces the Boltzmann entropy exactly. Boltzmann and Gibbs agree for the isolated (microcanonical) macrostate. *Verified: stationarity gives $p_i=e^{-1-\lambda}$ (constant in $i$ ) and $-\sum\frac1\Omega\ln\frac1\Omega=\ln\Omega$ , both checked symbolically in the verify script. Cf. Reif Ch. 3, Sethna Ch. 5.*

Limiting cases

Case	Result	Meaning
Classical continuum	$\Omega=\frac{1}{N!\,h^{3N}}\int_{H=E}\mathrm d^{3N}q\,\mathrm d^{3N}p$	microstates are phase-space points; the count is a volume divided by Planck cells and the Gibbs factor
Quantum discrete	$\Omega=\#\{\|n\rangle:E_n\in[E,E+\delta E]\}$	microstates are energy eigenstates; counting is genuinely discrete, indistinguishability built into the symmetrized Hilbert space
Two-state spins	$\Omega(N_\uparrow)=\binom{N}{N_\uparrow}$	a microstate is the full list of spin orientations; the macrostate fixes only total magnetization
Thermodynamic limit	$\tfrac1N\ln\Omega\to \text{finite}$	per-particle entropy converges; one macrostate dominates and fluctuations of intensive variables vanish as $N^{-1/2}$

Worked examples

1 · Counting a small system by hand, then trusting the formula (confidence builder)

Four labelled molecules occupy a box split into left and right halves. Enumerate the microstates, find the multiplicity of each macrostate $n=$ (number on the left), and compute the probability that all four end up on the same side.

Step 1. Total microstates: each of 4 molecules is independently left or right, so $2^4=\boxed{16}$ microstates.

Step 2. Multiplicities by macrostate, $\Omega(n)=\binom{4}{n}$ : $\Omega(0)=1,\ \Omega(1)=4,\ \Omega(2)=6,\ \Omega(3)=4,\ \Omega(4)=1$ . They sum to $1+4+6+4+16$ ? Check: $1+4+6+4+1=16$ ✓ — the uniform macrostate $n=2$ owns $6/16$ of everything.

Step 3. Probability all four on one side: $P(n=0)+P(n=4)=\frac{1}{16}+\frac{1}{16}=\frac{2}{16}=\boxed{1/8}$ .

Step 4. Scale up. For $N=100$ the same "all on one side" probability is $2\cdot(1/2)^{100}\approx1.6\times10^{-30}$ ; for $N=10^{23}$ it is $\sim10^{-3\times10^{22}}$ — a number with more zeros than there are atoms in the observable universe. The gas fills the box not by force but by counting.

2 · The width of equilibrium for a real mole (exam level)

A mole of gas ( $N=6.022\times10^{23}$ ) sits in a box. Estimate the typical fractional fluctuation of the number of molecules in the left half, and the entropy scale of the uniform macrostate versus a "10% denser on the left" macrostate.

Step 1. Fractional fluctuation from Eq. (8): $\dfrac{\sigma}{N/2}=\dfrac{1}{\sqrt N}=\dfrac{1}{\sqrt{6.022\times10^{23}}}=\boxed{1.3\times10^{-12}}$ . Density differences between the halves are spontaneously present only at the part-per-trillion level.

Step 2. How suppressed is a 10% imbalance? Put $n=0.55N$ into $\ln\Omega(n)\approx\ln\Omega(N/2)-\frac{2}{N}(n-N/2)^2$ . Here $(n-N/2)=0.05N$ , so the exponent is $-\frac{2}{N}(0.05N)^2=-0.005N$ .

Step 3. With $N=6.022\times10^{23}$ that is $-3.0\times10^{21}$ . The imbalanced macrostate is less probable by $e^{-3\times10^{21}}$ — utterly forbidden. In entropy units it lies $\Delta S=k_{\mathrm B}\cdot3.0\times10^{21}=4.2\times10^{-2}\ \mathrm{J/K}$ below equilibrium.

Step 4. Conclude: equilibrium is not a delicate balance but a mountain peak whose slopes are exponentially steep in $N$ . This is why the second law feels absolute even though the microscopic laws are reversible — you would wait longer than the age of the universe to see a mole spontaneously crowd 10% into one half.

Origins · 1 minLudwig Boltzmann and how this was discovered

Ludwig Boltzmann

1844–1906 · Vienna and Graz

Boltzmann spent his career insisting that the laws of heat are laws of counting — that entropy measures the number of microscopic ways a macrostate can be realized. His relation $S=k\ln W$ (in his notation $W$ , Wahrscheinlichkeit, "probability" or number of ways) unified thermodynamics with the atomic hypothesis, and it is carved on his headstone in Vienna. He faced fierce opposition from anti-atomists and from the reversibility and recurrence paradoxes of Loschmidt and Zermelo, which he answered with exactly the argument in this article: irreversibility is statistical, a matter of overwhelming probability, not microscopic law. He did not live to see atoms confirmed; the constant $k_{\mathrm B}$ that bears his name was first written down explicitly by Planck a few years later.

History

1738

— Daniel Bernoulli derives gas pressure from molecular collisions, the first kinetic model.

1860

— Maxwell publishes the distribution of molecular speeds, making the statistical description quantitative.

1872

— Boltzmann's H-theorem argues that a gas approaches equilibrium, provoking the reversibility debate.

1877

— Boltzmann links entropy to the number of microstates, the idea behind

S=k\ln W

1902

— Gibbs, in Elementary Principles in Statistical Mechanics, systematizes ensembles and coins "statistical mechanics" and "phase space."

Going further

Texts. Reif, Fundamentals of Statistical and Thermal Physics, Ch. 2–3 · Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 1–2 · Kardar, Statistical Physics of Particles, Ch. 4 · Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, Ch. 3 · Landau & Lifshitz, Statistical Physics Part 1, §1–7.

Provenance. All results are standard. A verifier should confirm: the binomial multiplicity $\Omega(n)=\binom{N}{n}$ , its peak at $n=N/2$ , and the variance $\sigma^2=N/4$ giving relative width $1/(2\sqrt N)$ (verify script, symbolic + numeric); the equivalence $S=k_{\mathrm B}\ln\Omega=\max[-k_{\mathrm B}\sum p_i\ln p_i]$ under normalization; the worked numbers ( $1/8$ for four molecules on one side; $1.3\times10^{-12}$ fractional fluctuation for a mole; the $e^{-3\times10^{21}}$ suppression of a 10% imbalance; Sackur–Tetrode $S/Nk\approx18.5$ for argon at STP as an order-of-magnitude anchor). The only fundamental constant used is $k_{\mathrm B}=1.380649\times10^{-23}\ \mathrm{J/K}$ (exact, 2019 SI).

Prerequisites. Phase Space · Hamiltonian mechanics. Leads to. Statistical Ensembles · Fundamental Postulate of Equal A Priori Probabilities · Accessible Phase Space and Energy Shells · Boltzmann Entropy · Phase Space Density and the Distribution Function · The Ergodic Hypothesis.

Isn't a "macrostate" just a fuzzy microstate — the same thing seen blurrily?

No — they are different kinds of object, and the difference is the whole point. A microstate is a single exact point $\Gamma$ in phase space. A macrostate is a set of microstates: all the $\Omega$ points sharing the same $(E,V,N)$ . Blurring one microstate would still leave you with one (imprecise) point; a macrostate is genuinely a huge collection. The map from microstate to macrostate is many-to-one, and the size of each preimage — the multiplicity $\Omega$ — is exactly what entropy measures.

Why the logarithm in S = k ln Ω? Why not just use Ω itself as the entropy?

Because entropy has to add for independent systems and $\Omega$ multiplies. Put two independent systems together and the joint microstate count is $\Omega=\Omega_1\Omega_2$ (any microstate of one with any of the other). We want their entropies to add, $S=S_1+S_2$ , matching thermodynamics where entropy is extensive. The only function turning products into sums is the logarithm, so $S=k_{\mathrm B}\ln\Omega$ is forced, not chosen. Boltzmann's constant $k_{\mathrm B}$ just fixes the units (joules per kelvin).

If every microstate is equally likely, why is the gas almost never all on one side?

Every microstate is equally likely — but macrostates are not, because they contain wildly different numbers of microstates. "All on one side" is a single microstate (per side); "uniform" is realized by $\binom{N}{N/2}$ microstates, an astronomically larger number. The uniform macrostate is favoured not because any arrangement is special but because it has vastly more arrangements to be in. Fair microstates, unfair macrostates.

How can entropy increase if Hamilton's equations are perfectly reversible?

This is Loschmidt's objection, and the resolution is that irreversibility is a statement about macrostates, not microstates. Every microscopic trajectory does have a valid time-reverse. But if you start in a low- $\Omega$ macrostate (gas in one corner), the overwhelming majority of accessible microstates lead to higher- $\Omega$ macrostates simply because those are exponentially more numerous. The reverse — spontaneously assembling into the corner — is allowed but has probability $\sim e^{-N}$ . The arrow comes from the special, low-entropy initial condition, not from the equations.

What exactly does the h³ᴺ do? We're doing classical mechanics — where does Planck's constant come from?

It sets the size of a "state." Classically, phase-space volume is a continuous real number with dimensions of (action) $^{3N}$ , so a raw volume can't be a count of states. Quantum mechanics tells us each independent state occupies a cell of volume $h^{3N}$ (one factor of $h$ per conjugate pair, from the uncertainty principle). Dividing by $h^{3N}$ turns the volume into a dimensionless integer that matches the quantum state count in the classical limit. It is the fingerprint of quantum mechanics surviving into the classical formulas.

Does the macrostate have to be (E, V, N)? What if I fix temperature instead?

You can fix any consistent set of macroscopic variables — that choice is exactly what defines an ensemble. Fixing $(E,V,N)$ for an isolated system gives the microcanonical ensemble, weighting microstates equally. Fixing $(T,V,N)$ (system in a heat bath) gives the canonical ensemble, weighting microstate $i$ by $e^{-E_i/k_{\mathrm B}T}$ . Fixing $(T,V,\mu)$ gives the grand canonical ensemble. All are probability distributions over the same microstates; they differ only in which macroscopic constraints are held fixed. That is the subject of Statistical Ensembles.

Is Ω ever actually computable, or is it just a formal symbol?

For a few clean models it is exact — the two-state spin system ( $\Omega=\binom{N}{n}$ ) and the ideal gas (whose energy-shell $\Omega$ gives the Sackur–Tetrode entropy, $S/Nk\approx18.5$ for argon at STP). For realistic interacting systems $\Omega$ is not computed directly; instead we switch to the canonical ensemble and compute a partition function, which is analytically or numerically far more tractable and yields the same thermodynamics in the thermodynamic limit. So $\Omega$ is the conceptual foundation, but the partition function is usually the working tool.