Statistical Mechanics

Bridging the Microscopic and Macroscopic Worlds

Statistical mechanics provides the microscopic foundation for thermodynamics, connecting the behavior of individual particles to macroscopic observables. It explains how the laws of thermodynamics emerge from the statistical behavior of large ensembles. Three ideas anchor the subject: macroscopic properties are statistical averages over microstates; ensembles give different statistical descriptions for different constraints; and phase transitions are collective phenomena governed by critical exponents and universality.

Explore Statistical Mechanics

Classical & Quantum Statistical Mechanics

Phase space and Liouville's theorem, the classical and quantum partition functions, Fermi-Dirac and Bose-Einstein statistics, ideal Fermi and Bose gases, and interacting systems (virial expansion, mean field theory).

Phase Transitions & Graduate Formalism

First- and second-order transitions, critical phenomena and universality, fluctuations, non-equilibrium statistical mechanics, and a graduate-level reference block (Keldysh, replica, bosonization, BCS, tensor networks).

From Microstates to Ensembles

The whole subject rests on a single, almost embarrassingly simple idea, and everything else is bookkeeping built on top of it. A microstate is a complete answer to the question “what is every degree of freedom doing right now?” — the position and momentum of every classical particle, or the occupation of every quantum level. A macrostate is the handful of numbers a thermometer or pressure gauge can actually read: energy $E$, volume $V$, particle number $N$, temperature $T$, pressure $P$. The defining feature of a many-body system is that a single macrostate is consistent with an astronomical number of microstates, and we have no way (and no need) to know which one the system currently occupies.

The fundamental postulate

Statistical mechanics resolves this ignorance with one assumption, the postulate of equal a priori probabilities:

For an isolated system in equilibrium, every microstate compatible with the imposed constraints (fixed $E$, $V$, $N$) is equally likely.

There is no preferred microstate; the system has no memory of how it got where it is, and over time its dynamics carries it democratically through all the states it is allowed to visit (the ergodic hypothesis makes this precise). Equal weighting is the least-biased choice — it assumes nothing beyond the constraints themselves — and from it every probability distribution in the theory is derived. If $\Omega$ accessible microstates are all equally probable, each carries probability $1/\Omega$, and the entropy that counts them is Boltzmann’s

\[S = k_B \ln \Omega.\]

Constraints select the ensemble

An ensemble is a conceptual collection of a huge number of imaginary copies of the system, each in a different microstate, distributed according to the probabilities the postulate dictates. Which ensemble we use is decided entirely by what the system is allowed to exchange with its surroundings — that is, by the physical boundary conditions:

Physical situation	What’s held fixed	What fluctuates	Ensemble	Natural variables
Isolated — insulated, rigid, sealed	$E$, $V$, $N$	nothing	Microcanonical	$N, V, E$
Closed, in a heat bath — diathermal walls	$T$, $V$, $N$	energy $E$	Canonical	$N, V, T$
Open, in a reservoir — permeable walls	$T$, $V$, $\mu$	energy $E$ and number $N$	Grand canonical	$\mu, V, T$

The logic is uniform. Start from a large isolated “system + reservoir” whose total $(E, V, N)$ is fixed, and apply the fundamental postulate to the combined system. Then ask only about the small subsystem of interest. Summing over the reservoir’s many microstates trades a fixed conserved quantity for the intensive variable conjugate to it:

Open the walls to energy flow and the subsystem’s energy is no longer fixed; the reservoir’s vast heat capacity pins the shared temperature $T$ instead. Counting reservoir states produces the Boltzmann weight $e^{-\beta E_i}$ with $\beta = 1/(k_B T)$, and the microcanonical $\Omega$ becomes the canonical partition function $Z$.
Open the walls to particle flow as well and $N$ also floats; the reservoir now fixes the chemical potential $\mu$, attaching a factor $e^{\beta \mu N}$ to each state and turning $Z$ into the grand partition function $\mathcal{Z}$.

Each relaxation swaps a held-fixed extensive variable ($E$, then $N$) for its conjugate intensive partner ($T$, then $\mu$) — exactly the Legendre-transform structure that takes the entropy $S(E,V,N)$ to the Helmholtz free energy $F(T,V,N)$ to the grand potential $\Omega(T,V,\mu)$ in thermodynamics. The microstates are always counted the same way; only the constraint, and hence the statistical weight, changes.

Why it doesn’t matter which one you pick

These three descriptions are not competing theories — they are the same physics viewed through different boundary conditions, and in the thermodynamic limit ($N \to \infty$ at fixed densities) they give identical predictions for every macroscopic observable. The reason is that relative fluctuations shrink as $1/\sqrt{N}$: in the canonical ensemble the energy is technically allowed to vary, but for $N \sim 10^{23}$ it is overwhelmingly likely to sit within a part in $10^{11}$ of its mean, so “fixed $E$” and “fixed $T$” describe the same equilibrium. Ensemble equivalence means you are free to choose whichever ensemble makes the mathematics easiest — almost always the canonical one, because summing the unconstrained $e^{-\beta E_i}$ over all states is far simpler than counting only those states with one exact energy. The sections below build out each ensemble in turn and then put the partition function to work.

Fundamental Principles

Microstates and Macrostates

A microstate is a complete specification of the quantum state of every particle; a macrostate is a specification of the macroscopic variables ($T, P, V, N, E$). The two diagrams below contrast them.

Why counting microstates yields thermodynamics. Imagine flipping 100 coins. Every specific sequence is equally likely, yet you almost always see close to 50 heads — there are astronomically more ways to arrange “about half heads” than “all heads.” A gas of $10^{23}$ particles takes this to the extreme: the overwhelming majority of microstates look macroscopically identical (uniform density, a single temperature), so the system is found in that macrostate with near-certainty. Entropy $S = k_B \ln \Omega$ is just the logarithm of how many microstates wear a given macroscopic face, and the Second Law becomes a near-tautology — systems drift toward macrostates that more microstates correspond to. Thermodynamics is what statistics looks like when the numbers are enormous. The tool that makes the counting tractable is the partition function, introduced next through the ensembles.

Statistical Ensembles

Microcanonical Ensemble (NVE)

Isolated system with fixed energy, volume, and particle number

Partition function: $\Omega(E,V,N)$ = number of microstates

Entropy: $S = k_B \ln \Omega$

Canonical Ensemble (NVT)

System in thermal equilibrium with heat bath at temperature T

Partition function:

$Z = \sum_i e^{-\beta E_i} = \text{Tr}(e^{-\beta H})$

Where $\beta = \frac{1}{k_B T}$

Helmholtz free energy: $F = -k_B T \ln Z$

Grand Canonical Ensemble (μVT)

System can exchange particles and energy with reservoir

Grand partition function:

$\mathcal{Z} = \sum_{N=0}^{\infty} \sum_i e^{-\beta(E_i - \mu N)}$

Grand potential: $\Omega = -k_B T \ln \mathcal{Z}$

Key Takeaways

Entropy counts microstates. Boltzmann’s $S = k_B \ln \Omega$ links microscopic configurations to macroscopic thermodynamics.
The partition function is everything. From $Z$ you derive all thermodynamics: free energy, entropy, energy, and response functions.
Ensembles agree at large $N$. Microcanonical, canonical, and grand canonical descriptions become equivalent in the thermodynamic limit.
Quantum statistics matter. Bosons (Bose-Einstein) and fermions (Fermi-Dirac) behave radically differently at low temperature.
Phase transitions are collective. Singularities in $Z$ emerge only in the thermodynamic limit; universality groups them by symmetry and dimension.
Fluctuations encode response. The fluctuation-dissipation theorem connects equilibrium fluctuations to how a system responds to perturbation.