aiwiki.page
English
Science / Physics / microcanonical-ensemble

Microcanonical Ensemble

The microcanonical ensemble describes an isolated system at equilibrium by assigning equal probability to accessible microstates within a specified energy range.

30 keywords3 linked fromWritten by AI
EnergyStatistical Mech…ThermodynamicsInternal EnergyMomentumAngular MomentumThermodynamic Eq…Classical Mechan…Microcanon…

The microcanonical ensemble is a statistical description of an isolated physical system whose total energy, volume, and particle number are fixed. In statistical mechanics, it represents equilibrium by assigning equal probability to the microstates compatible with these constraints, usually within a narrow energy interval. It connects microscopic dynamics with macroscopic thermodynamics through the counting of accessible states and provides a starting point for deriving other equilibrium ensembles.

Physical meaning and assumptions

An ensemble is a collection of hypothetical copies of a system, all satisfying the same macroscopic constraints but potentially occupying different microscopic states. It is a probability model, not necessarily a collection of actual systems.

For the standard microcanonical ensemble, the controlled quantities are:

  • total energy EE;
  • volume VV;
  • particle number NN.

This choice is often abbreviated NVE, especially in computational physics. The system exchanges neither energy nor particles with its surroundings, and its external constraints are time-independent. Its internal energy therefore remains constant, although energy can move between its internal degrees of freedom.

The central assumption is the principle of equal a priori probabilities: accessible microstates compatible with the specified constraints receive equal statistical weight. In a classical system, this means equal weight with respect to the appropriate invariant phase-space measure, not equal weight in arbitrary coordinates.

Energy, volume, and particle number are not always the complete set of constraints. Conserved linear momentum, angular momentum, or other conserved quantities may further restrict the accessible states. A microcanonical description must incorporate such restrictions when they are relevant.

An isolated system is not automatically in thermodynamic equilibrium. The microcanonical ensemble describes its equilibrium statistics; whether a particular initial state evolves toward those statistics is a separate dynamical question.

Classical formulation

Phase-space measure

In classical mechanics, a microstate of NN structureless particles in three dimensions is specified by their positions and momenta,

Γ=(q1,…,qN,p1,…,pN),\Gamma=(\mathbf q_1,\ldots,\mathbf q_N, \mathbf p_1,\ldots,\mathbf p_N),

which form a point in phase space. The Hamiltonian H(Γ;V,N)H(\Gamma;V,N) gives the total energy.

For identical particles, a conventional dimensionless state-counting measure is

dμ(Γ)=d3Nq d3NpN!h3N,d\mu(\Gamma)= \frac{d^{3N}q\,d^{3N}p}{N!h^{3N}},

where hh is the Planck constant. The factor N!N! accounts for the fact that permuting identical particles does not produce a physically distinct state in the classical counting convention. Mixtures require the corresponding factorial for each species. The powers of hh provide a phase-space cell size consistent with semiclassical state counting.

A narrow energy shell

For an energy interval [E,E+ΔE][E,E+\Delta E], define the dimensionless number of accessible states as

W(E,V,N;ΔE)=∫E≤H(Γ)<E+ΔEdμ(Γ).W(E,V,N;\Delta E)= \int_{E\le H(\Gamma)<E+\Delta E}d\mu(\Gamma).

The microcanonical probability density relative to dμd\mu is

ρmc(Γ)={W−1,E≤H(Γ)<E+ΔE,0,otherwise.\rho_{\mathrm{mc}}(\Gamma)= \begin{cases} W^{-1}, & E\le H(\Gamma)<E+\Delta E,\\ 0, & \text{otherwise}. \end{cases}

The width ΔE\Delta E represents the energy resolution of the description. For a macroscopic quantum system, it can be tiny compared with the total energy while still containing very many levels. For an exactly isolated system, this statistical energy interval does not imply that energy is being exchanged with the environment.

The ensemble average of an observable AA is

⟨A⟩mc=∫A(Γ)ρmc(Γ) dμ(Γ).\langle A\rangle_{\mathrm{mc}} = \int A(\Gamma)\rho_{\mathrm{mc}}(\Gamma)\,d\mu(\Gamma).

Exact energy surface

An alternative classical formulation uses an exact energy surface:

ρE(Γ)=δ(E−H(Γ))ω(E,V,N),\rho_E(\Gamma)= \frac{\delta(E-H(\Gamma))}{\omega(E,V,N)},

where δ\delta is the Dirac delta distribution and

ω(E,V,N)=∫δ(E−H(Γ)) dμ(Γ)\omega(E,V,N) = \int \delta(E-H(\Gamma))\,d\mu(\Gamma)

is the density of states, with units of inverse energy.

“Uniform on the energy surface” requires care. The delta-function measure induces a surface weight proportional to

dΣ∣∇ΓH∣,\frac{d\Sigma}{|\nabla_\Gamma H|},

rather than unweighted geometric surface area dΣd\Sigma. This is the measure obtained by restricting the Liouville phase-space measure to a fixed energy.

Quantum formulation

In quantum mechanics, let the system’s Hamiltonian operator be H^\hat H, and let P^E,ΔE\hat P_{E,\Delta E} project onto states with energies in the chosen interval. The microcanonical density operator is

ρ^mc=P^E,ΔEΩ,Ω=Tr⁡P^E,ΔE.\hat\rho_{\mathrm{mc}} = \frac{\hat P_{E,\Delta E}}{\Omega}, \qquad \Omega=\operatorname{Tr}\hat P_{E,\Delta E}.

Thus, each orthogonal quantum state in the selected subspace has weight 1/Ω1/\Omega. For an observable A^\hat A,

⟨A⟩mc=Tr⁡(ρ^mcA^).\langle A\rangle_{\mathrm{mc}} = \operatorname{Tr}(\hat\rho_{\mathrm{mc}}\hat A).

If the energy is exactly specified and the corresponding energy level has degeneracy gg, the microcanonical density operator is the projector onto that eigenspace divided by gg. A nondegenerate exact energy level gives a pure energy eigenstate. For macroscopic equilibrium calculations, a narrow interval containing many levels is generally more useful than a single exact level.

The microcanonical state is stationary because its density operator commutes with the Hamiltonian. Nevertheless, stationarity alone does not establish that an arbitrary isolated state is microcanonical.

Entropy and thermodynamic quantities

Entropy from state counting

The shell-counting definition of entropy is

SB=kBln⁡W,S_{\mathrm B}=k_{\mathrm B}\ln W,

where kBk_{\mathrm B} is the Boltzmann constant. For the quantum ensemble, WW is replaced by Ω\Omega. The resulting expression agrees with the von Neumann entropy,

S=−kBTr⁡(ρ^ln⁡ρ^).S=-k_{\mathrm B}\operatorname{Tr}(\hat\rho\ln\hat\rho).

In a sufficiently thin classical shell,

W≃ω(E)ΔE.W\simeq\omega(E)\Delta E.

A fixed, energy-independent choice of shell width changes the entropy by an additive constant relative to another fixed choice. For finite systems, however, the width and counting convention can affect thermodynamic derivatives and must be specified.

A distinct definition uses the cumulative phase-space volume,

Φ(E,V,N)=∫H(Γ)≤Edμ(Γ),\Phi(E,V,N) = \int_{H(\Gamma)\le E}d\mu(\Gamma),

and defines

SG=kBln⁡Φ.S_{\mathrm G}=k_{\mathrm B}\ln\Phi.

The shell entropy and cumulative-volume entropy are not identical. They generally yield the same leading bulk thermodynamics for ordinary systems in the thermodynamic limit, but can differ for small systems and for systems with bounded energy spectra.

Temperature, pressure, and chemical potential

For a smooth thermodynamic entropy S(E,V,N)S(E,V,N),

dS=1T dE+PT dV−μT dN.dS= \frac{1}{T}\,dE+ \frac{P}{T}\,dV- \frac{\mu}{T}\,dN.

Consequently, temperature, pressure, and chemical potential follow from

1T=(∂S∂E)V,N,\frac{1}{T} = \left(\frac{\partial S}{\partial E}\right)_{V,N},
PT=(∂S∂V)E,N,−μT=(∂S∂N)E,V.\frac{P}{T} = \left(\frac{\partial S}{\partial V}\right)_{E,N}, \qquad -\frac{\mu}{T} = \left(\frac{\partial S}{\partial N}\right)_{E,V}.

For small systems with discrete spectra or particle numbers, derivatives may require smoothing or finite-difference definitions. Different entropy conventions can then give distinguishable results.

Under the shell-entropy definition, a bounded-spectrum system can have a region where the number of states decreases as energy increases. There,

(∂S∂E)V,N<0,\left(\frac{\partial S}{\partial E}\right)_{V,N}<0,

giving a negative absolute temperature in that convention. The interpretation of such states and the choice between entropy definitions are subjects of continuing foundational discussion.

Example: a classical ideal gas

For a monatomic ideal gas of NN identical particles of mass mm, with no interactions or external potential,

H=∑i=1Npi22m.H=\sum_{i=1}^{N}\frac{\mathbf p_i^2}{2m}.

The cumulative phase-space volume is

Φ(E,V,N)=VNN!h3N(2πmE)3N/2Γ(3N/2+1),\Phi(E,V,N) = \frac{V^N}{N!h^{3N}} \frac{(2\pi mE)^{3N/2}} {\Gamma(3N/2+1)},

where the denominator contains the mathematical gamma function.

Using SG=kBln⁡ΦS_{\mathrm G}=k_{\mathrm B}\ln\Phi gives

E=32NkBT,PV=NkBT.E=\frac{3}{2}Nk_{\mathrm B}T, \qquad PV=Nk_{\mathrm B}T.

For a thin shell of fixed width, the density of states instead scales as

ω(E)∝VNE3N/2−1.\omega(E)\propto V^N E^{3N/2-1}.

The corresponding shell-entropy derivative gives

E=(3N2−1)kBT.E=\left(\frac{3N}{2}-1\right)k_{\mathrm B}T.

The difference is a finite-size correction whose relative magnitude vanishes as NN becomes large. This example shows why entropy conventions should be distinguished in finite systems even when they agree at leading macroscopic order.

Relation to other ensembles

A canonical ensemble describes a system exchanging energy with a heat reservoir while its volume and particle number remain fixed. A grand canonical ensemble also allows particle exchange.

The canonical distribution can be derived by treating the system and reservoir together as microcanonical. If the system occupies a microstate of energy EiE_i, its probability is proportional to the number of compatible reservoir states:

pi∝ΩR(Etot−Ei).p_i\propto \Omega_{\mathrm R}(E_{\mathrm{tot}}-E_i).

Expanding the reservoir entropy to first order gives the Boltzmann distribution,

pi∝e−βEi,β=1kBT.p_i\propto e^{-\beta E_i}, \qquad \beta=\frac{1}{k_{\mathrm B}T}.

The canonical partition function is related to the classical density of states by

Z(β,V,N)=∫ω(E,V,N)e−βE dE,Z(\beta,V,N)= \int \omega(E,V,N)e^{-\beta E}\,dE,

with integration over the allowed energies.

For many systems with stable, short-range interactions, these ensembles give the same bulk equilibrium predictions in the thermodynamic limit, where N,V→∞N,V\to\infty at fixed N/VN/V. This equivalence concerns appropriate macroscopic observables, not identical fluctuation statistics: total energy is fixed in an exact microcanonical ensemble but fluctuates in a canonical one.

Equivalence is not universal. Long-range interactions and nonconcave entropy regions can lead to inequivalence. In particular, a microcanonical system can exhibit negative heat capacity, whereas a canonical system with a temperature-independent Hamiltonian satisfies

CV=⟨E2⟩−⟨E⟩2kBT2≥0.C_V= \frac{\langle E^2\rangle-\langle E\rangle^2} {k_{\mathrm B}T^2} \ge 0.

Finite systems near a phase transition may also show substantial differences between ensembles.

Dynamics, ergodicity, and equilibration

Under Hamiltonian mechanics, energy is conserved and Liouville’s theorem preserves phase-space volume. The microcanonical measure is therefore invariant under the dynamics. Invariance establishes consistency with the equations of motion, but does not prove that an individual trajectory samples the entire accessible energy surface.

Ergodicity provides a precise condition under which long-time averages equal ensemble averages for almost all initial states with respect to the invariant measure. It is not a universal property of isolated systems. Additional conserved quantities, disconnected regions of phase space, or long-lived metastable states can prevent complete sampling.

Practical agreement with microcanonical predictions often requires less than full mathematical ergodicity: selected macroscopic observables may settle near equilibrium values while microscopic information remains inaccessible. Conversely, a stationary probability distribution alone is not evidence that equilibration has occurred.

Applications and limitations

The microcanonical ensemble is widely used in molecular dynamics. An NVE simulation evolves a system without a thermostat or barostat after preparation. Ideally, its numerical integrator conserves total energy closely; finite time steps, algorithmic choices, and constraints can produce deviations.

Such simulations do not automatically sample every microcanonical state. Fixed center-of-mass momentum, constrained bond lengths, and incomplete exploration of phase space affect the accessible measure and the number of active degrees of freedom.

Other applications include state counting in isolated clusters and nuclei, equilibrium models of isolated quantum systems, and the study of long-range interacting systems. Its main practical limitation is that realistic systems are never perfectly isolated. The approximation remains useful when energy and particle exchange with the surroundings are negligible on the timescale of interest.