Fundamental Postulate of Equal A Priori Probabilities

The fundamental postulate is not a museum piece — it is the operating assumption of every equilibrium simulation run today. Microcanonical molecular dynamics assumes that a single trajectory ergodically samples the energy shell, and Metropolis Monte Carlo is deliberately engineered to converge to an ensemble derived from it; when either "fails to equilibrate," what has failed is precisely effective ergodicity — the simulation analogue of the broken-ergodicity caveat this article insists on. Diagnosing and fixing that failure (replica exchange, enhanced sampling) is a central craft of computational physics. The postulate's sharpest modern payoff is negative absolute temperature. For a system whose energy spectrum is bounded above — nuclear spins in a field, or ultracold atoms in an optical lattice — the count $\Omega(E)$ rises, peaks, and falls, so beyond the peak $\beta=\partial\ln\Omega/\partial E$ turns negative. A population-inverted spin system is genuinely hotter than infinite temperature, a counterintuitive prediction that follows directly from counting accessible microstates and that has been realized in the laboratory (Purcell and Pound in 1951 for nuclear spins; a motional negative-temperature state of cold atoms in 2013). Whenever you read that some exotic system has "negative temperature," you are reading a direct consequence of the equal-a-priori-probability postulate applied to a bounded spectrum.

Shake a sealed box of $10^{23}$ molecules with a fixed total energy and let it settle. After equilibration you have no way to know which exact microscopic configuration it landed in — and crucially, no configuration of the right energy is special. Faced with total ignorance beyond the fixed energy, the only unbiased bet is to weight every accessible configuration equally. That democratic bet is the fundamental postulate, and its entire justification is the honest confession that you know nothing more than the macrostate. Dynamics earns the postulate rather than merely asserting it. A chaotic Hamiltonian flow stretches and folds any initial blob of phase space until it is smeared uniformly across the energy shell — the mixing property. Liouville's theorem forbids the blob from shrinking or growing, so the only thing it can do is spread evenly, which is precisely the uniform distribution. The postulate is thus not an arbitrary guess but the fixed point that realistic dynamics converges to. There is a third, equivalent way to see it: the uniform distribution is maximal ignorance with a constraint. Among all probability distributions supported on the energy shell, the uniform one has the largest Gibbs entropy $-k_B\int\rho\ln\rho$. Any departure from uniformity would be information — a claim to know that some microstates are more likely — and for an isolated system in equilibrium you have no such information. Ignorance, stationarity, and dynamics all point to the same distribution.

The three arrows all land on the same shaded shell: ignorance, maximum entropy, and mixing dynamics (with Liouville guaranteeing stationarity) each independently select the uniform distribution. That over-determination is why the postulate feels inevitable even though it cannot be derived from mechanics alone.

This second figure is the reason "most probable" becomes "certain." Because the number of microstates grows like $e^{N\sigma}$, the joint count of two systems sharing energy is a spike of relative width $1/\sqrt N$. For a macroscopic $N$ the spike is so narrow that the equilibrium energy split $E_1^*$ is realized with overwhelming probability — the average and the most probable value coincide to twelve decimal places.

The postulate has a crisp discrete statement: probability $1/\Omega$ on each of the $\Omega$ accessible microstates, zero on everything else. Everything downstream is a consequence of this one line.

where: $\nu$ — a microstate; $H(\nu)$ — its energy; $E,\Delta$ — the energy and the (thin) shell width fixing the macrostate; $\Omega$ — the number of accessible microstates in the shell. The "otherwise = 0" is as important as the "$1/\Omega$": microstates of the wrong energy, or that the dynamics cannot reach, get no weight. In continuous phase space the same idea is a density concentrated on the energy surface, normalized by the density of states.

where: $\rho$ — the microcanonical phase-space density; $\delta$ — the Dirac delta pinning support to the surface $H=E$; $h^{3N}$ — Planck's constant per degree of freedom, making $\Omega$ dimensionless; $N!$ — the Gibbs factor correcting for indistinguishable particles. The $h^{3N}N!$ does not affect equiprobability, but it fixes the absolute entropy and resolves the Gibbs paradox — without it entropy would not be extensive. The single most consequential output is that the entropy of this uniform distribution is Boltzmann's logarithm of a count — the bridge from combinatorics to thermodynamics.

where: $S$ — thermodynamic entropy; $k_B$ — Boltzmann's constant (exact since the 2019 SI redefinition). Read physically: entropy simply counts accessible microstates on a logarithmic scale, so more accessible phase space means more entropy — geometry becomes thermodynamics. Once $S(E,V,N)$ is known, temperature, pressure, and chemical potential are just its derivatives — thermodynamics falls out by differentiation, with no further assumptions.

where: $T,p,\mu$ — temperature, pressure, chemical potential. Each intensive variable is the rate at which the log-count of microstates changes as you add energy, volume, or a particle — temperature, for instance, measures how eager the system is to accept energy. Finally, Liouville's theorem is what makes the uniform density a legitimate equilibrium distribution: any density depending on phase-space point only through $H$ is stationary.

where: $\{\cdot,\cdot\}$ — the Poisson bracket; $\rho(H)$ — any function of the Hamiltonian. Because $\{\rho(H),H\}=\rho'(H)\{H,H\}=0$, a shell-uniform density never changes in time — the postulate describes a genuine steady state, not a transient.

The postulate's validity depends on whether the dynamics actually explores the shell; its form changes when extra conservation laws or trapping intervene:

1 · Counting the equilibrium macrostate of a coin gas (confidence builder)

A toy "gas" is $N$ independent two-state units, each up or down, at fixed total energy corresponding to exactly $N/2$ up. Treat every arrangement as an equiprobable microstate. For $N=4$ enumerate the microstates of the equilibrium macrostate; then for $N=1000$ estimate how sharply the system sits at $n_\uparrow=N/2$.

Step 1. Enumerate for $N=4$. The macrostate "$n_\uparrow=2$" is realized by $\binom{4}{2}=\mathbf{6}$ microstates: ↑↑↓↓, ↑↓↑↓, ↑↓↓↑, ↓↑↑↓, ↓↑↓↑, ↓↓↑↑. Compare the extreme macrostate "$n_\uparrow=4$" (all up), realized by only $\binom{4}{4}=1$. Even at $N=4$ the balanced macrostate is six times more populated.

Step 2. The postulate assigns each microstate $1/\Omega$. With $\Omega_{\text{total}}=2^4=16$, the probability of the $n_\uparrow=2$ macrostate is $6/16=\mathbf{0.375}$, versus $1/16=0.0625$ for all-up. The equal-weight-per-microstate rule makes the balanced macrostate the likeliest purely because it has the most microstates.

Step 3. Sharpen for $N=1000$. The multiplicity $\binom{N}{n_\uparrow}$ is a Gaussian in $n_\uparrow$ centered at $N/2$ with standard deviation $\sigma=\tfrac12\sqrt N$. The relative width is $\dfrac{\sigma}{N/2}=\dfrac{1}{\sqrt N}=\dfrac{1}{\sqrt{1000}}=\mathbf{0.0316}$ — the macrostate is pinned to within about 3% of $n_\uparrow=N/2$.

Step 4. Scale to a real gas. For $N\sim10^{24}$ the relative width is $1/\sqrt{10^{24}}=\mathbf{10^{-12}}$. The equilibrium macrostate is realized to twelve decimal places; deviations are utterly unobservable. This is why "most probable" becomes "certain," and why equal weighting of microstates reproduces sharp, reproducible thermodynamics.

2 · Deriving the zeroth law by counting for two ideal gases (exam level)

Two monatomic ideal gases with $\Omega_i(E_i)\propto E_i^{3N_i/2}$ are placed in thermal contact at fixed total energy $E$. Use equal a priori probability to find the equilibrium energy split and show it means $T_1=T_2$.

Step 1. Write the joint count. By equiprobability the probability of split $E_1$ is $\propto\Omega_1(E_1)\Omega_2(E-E_1)\propto E_1^{3N_1/2}(E-E_1)^{3N_2/2}$. Every joint microstate is equally likely, so the most-populated split dominates.

Step 2. Maximize the log. Setting $\partial_{E_1}[\tfrac{3N_1}{2}\ln E_1+\tfrac{3N_2}{2}\ln(E-E_1)]=0$ gives

Step 3. Read off the split. Solving, $E_1^*=\dfrac{N_1}{N_1+N_2}E$: energy is shared in proportion to particle number. For equal gases $N_1=N_2$ this is $E_1^*=E/2$ — equal energy each.

Step 4. Identify the temperature. The slope $\beta_i=\partial\ln\Omega_i/\partial E_i=\tfrac{3N_i}{2E_i}$ equals $1/(k_BT_i)$, so the equilibrium condition $\beta_1=\beta_2$ is exactly $T_1=T_2$. From $\tfrac{3N_i}{2E_i}=\tfrac{1}{k_BT_i}$ we even recover the equipartition result $E_i=\tfrac{3}{2}N_ik_BT_i$ — three-halves $k_BT$ per particle — from counting alone.

Step 5. Conclude. Two systems free to exchange energy settle at equal temperature, and temperature itself is defined as $1/(k_B\partial\ln\Omega/\partial E)$. The zeroth law is not an independent postulate; it is a theorem about which energy split has the most microstates, made certain by the $1/\sqrt N$ sharpness of the previous example.

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

Provenance. Structure follows Pathria & Beale ch. 1–3, Kardar ch. 4, Reif ch. 2–3, Sethna ch. 3, Landau–Lifshitz §1–4. A verifier should recompute: (i) the MaxEnt derivation yielding $P_\nu=1/\Omega$ by Lagrange multipliers and the second-derivative maximum test; (ii) the Boltzmann formula $S=k_B\ln\Omega$ as the Gibbs entropy of the uniform distribution; (iii) the thermal-contact condition $\partial\ln\Omega_1/\partial E_1=\partial\ln\Omega_2/\partial E_2$ with $\beta=1/(k_BT)$, and the ideal-gas split $E_1^*=N_1E/(N_1+N_2)$; (iv) the $\binom{4}{2}=6$ and $1/\sqrt N$ counting numbers. Constant: $k_B=1.380649\times10^{-23}\,\mathrm{J\,K^{-1}}$ (exact, SI 2019). All numbers are reproduced in the companion verification script.

Prerequisites. Microstates and macrostates · phase space · accessible phase space and energy shells · Liouville's theorem. Leads to. Statistical ensembles · Boltzmann entropy · the ergodic hypothesis · Gibbs entropy.