Boltzmann Distribution

The Boltzmann distribution (also known as the canonical distribution or Gibbs distribution) is the fundamental probability distribution of equilibrium statistical mechanics that governs how microscopic particles or macroscopic systems distribute across available energy states when coupled to a thermal reservoir at absolute temperature T. First derived for classical gas molecules by Austrian physicist Ludwig Boltzmann between 1868 and 1877 and generalized to arbitrary interacting ensembles by American physicist Josiah Willard Gibbs in 1902, the distribution states that the probability p_i of finding a system in a specific discrete microscopic state (microstate) i with energy E_i decreases as a Boltzmann exponential function exp(−*E*_i / k_BT) of the ratio of state energy to thermal energy k_BT:

p_i = (1 / Z) exp(−*E*_i / [k_BT]) = e^−β*E*_i / ∑_j e^−β*E*_j

where k_B = 1.380649 × 10⁻²³ J K⁻¹ is the Boltzmann constant, β = 1 / (k_BT) is the thermodynamic beta (inverse thermal energy), and Z = ∑_j exp(−*E*_j / k_BT) is the canonical partition function summed over all accessible microstates j. When multiple distinct microstates share an identical energy level E_n—a property known as quantum or classical degeneracy g_n—the probability of observing the system with energy E_n is given by P(E_n) = (g_n / Z) exp(−*E*_n / k_BT), where Z = ∑_n g_n exp(−*E*_n / k_BT).

The dimensionless quantity exp(−*E*_i / k_BT) is known as the Boltzmann factor. Taking the ratio of probabilities for any two microstates i and j eliminates the partition function Z, yielding the universal relative population ratio p_i / p_j = exp(−[E_i − E_j] / k_BT). This single relation bridges microscopic quantum mechanics and macroscopic thermodynamics: it governs the population of atomic and vibrational energy levels in spectroscopy, the ideal gas law and Maxwell–Boltzmann velocity distribution, the chemical activation barrier in the Arrhenius equation, voltage-gated ion channel opening in neurobiology, and two-state protein folding equilibria via Gibbs free energy.

Master Statistical Mechanics: Test your intuition for Boltzmann distribution laws, partition functions, and thermodynamic equations in EquaGuess.

System-Reservoir Coupling in the Canonical Ensemble

To derive the Boltzmann distribution from the fundamental postulate of statistical mechanics—which states that an isolated composite system in equilibrium is equally likely to occupy any of its accessible microstates—consider a small system S in weak thermal contact with a heat reservoir R at temperature T. The combined system S + R is thermally and mechanically isolated with constant total energy E_total = E_i + E_R, where E_i ≪ E_total is the energy of system S when it occupies a single specified microstate i, and E_R = E_total − E_i is the remaining energy in the reservoir.

Because the composite system is isolated, the probability p_i that system S occupies microstate i is directly proportional to the multiplicity Ω_R(E_total − E_i)—the total number of microstates available to the reservoir when its energy is E_total − E_i. Because Ω_R scales multiplicatively with system size, it is convenient to expand its natural logarithm using Boltzmann's definition of entropy, S_R(E_R) = k_B ln Ω_R(E_R), in a first-order Taylor series around E_total:

ln Ω_R(E_total − E_i) = S_R(E_total − E_i) / k_B ≈ [S_R(E_total) / k_B] − (1 / k_B)(∂*S*_R / ∂*E*_R)_V,N E_i

By the fundamental thermodynamic identity d*E* = T d*S* − P d*V* + μ d*N*, the partial derivative of reservoir entropy with respect to energy at constant volume and particle number defines the reciprocal of absolute temperature: (∂*S*_R / ∂*E*_R)_V,N = 1 / T. Higher-order quadratic derivatives ∂²S_R / ∂*E*_R² = −1 / (T²C_V) vanish in the limit of a large reservoir with heat capacity C_V → ∞. Exponentiating both sides yields:

Ω_R(E_total − E_i) = Ω_R(E_total) exp(−*E*_i / [k_BT])

Normalizing the probabilities over all microstates so that ∑_i p_i = 1 introduces the canonical partition function Z = ∑_i exp(−*E*_i / k_BT) as the normalization denominator, establishing p_i = Z⁻¹ exp(−*E*_i / k_BT).

Maximum Entropy Derivation via Lagrange Multipliers

An independent information-theoretic derivation, formulated by Gibbs (1902) and Edwin T. Jaynes (1957), obtains the Boltzmann distribution by maximizing the Gibbs–Shannon entropy functional S = −*k*_B ∑_i p_i ln p_i subject to two macroscopic constraints: probability normalization (∑_i p_i = 1) and a fixed ensemble average internal energy (⟨*E*⟩ = ∑_i p_iE_i = U). Introducing Lagrange multipliers α and β to construct the constrained variational functional L({*p*_i}) and setting ∂*L* / ∂*p*_i = 0 for each microstate i gives:

(∂ / ∂*p*_i)[−*k*_B ∑_j p_j ln p_j − α*k*_B(∑_j p_j − 1) − β*k*_B(∑_j p_jE_j − U)] = −*k*_B(ln p_i + 1 + α + β*E*_i) = 0

Solving for p_i yields the exponential distribution p_i = exp(−1 − α) exp(−β*E*_i) = Z⁻¹ exp(−β*E*_i), where matching the thermodynamic relation ∂*S* / ∂*U* = 1 / T identifies β = 1 / (k_BT). Thus, the Boltzmann distribution represents the macroscopically unbiased probability assignment that maximizes microscopic uncertainty for a given mean thermal energy U.

Once the canonical partition function Z(T, V, N) = ∑_i exp(−β*E*_i) is evaluated for a physical system, every macroscopic equilibrium thermodynamic property follows directly by differentiation of ln Z. The ensemble mean internal energy U = ⟨*E*⟩ is obtained by taking the negative derivative of ln Z with respect to β:

U = ⟨*E*⟩ = ∑_i p_iE_i = (1 / Z) ∑_i E_i e^−β*E*_i = −(∂ ln Z / ∂β)_V,N = k_BT² (∂ ln Z / ∂*T*)_V,N

Similarly, the variance of energy fluctuations around the mean is given by the second derivative ∂²(ln Z) / ∂β² = ⟨*E*²⟩ − ⟨*E*⟩² = k_BT²C_V, proving that the heat capacity at constant volume C_V is positive and proportional to mean-square thermal energy fluctuations. Substituting p_i = Z⁻¹ exp(−β*E*_i) into the Gibbs entropy formula S = −*k*_B ∑_i p_i ln p_i yields S = k_B(ln Z + β*U*) = U / T + k_B ln Z. Rearranging this equation for the Helmholtz free energy F = U − TS establishes the bridge equation of canonical statistical mechanics:

F = U − TS = −*k*_BT ln Z

From the differential of Helmholtz free energy d*F* = −*S* d*T* − P d*V* + μ d*N*, macroscopic pressure is P = −(∂*F* / ∂*V*)_T,N = k_BT(∂ ln Z / ∂*V*)_T,N, and chemical potential is μ = (∂*F* / ∂*N*)_T,V = −*k*_BT(∂ ln Z / ∂*N*)_T,V.

Maxwell–Boltzmann Speed Distribution and the Barometric Formula

Applying the Boltzmann distribution to the continuous phase space of an ideal gas of particles of mass m yields both spatial density and velocity distributions. In a uniform gravitational field g, the potential energy of a molecule at altitude z is E_pot(z) = mgz; the Boltzmann factor immediately gives the barometric exponential decay of particle number density with height: n(z) = n(0) exp(−*mgz* / k_BT).

Simultaneously, the translational kinetic energy of a molecule moving with velocity vector v = (v_x, v_y, v_z) is E_trans = ½*m*(v_x² + v_y² + v_z²). Because the three velocity components are independent in Cartesian momentum space, the probability density for a single velocity component v_x is a Gaussian normal distribution with variance σ² = k_BT / m. Integrating over all directions in spherical velocity coordinates—where a spherical shell of thickness d*v* at speed v has phase-space volume (degeneracy weight) 4π*v*² d*v*—produces the Maxwell–Boltzmann speed distribution:

f(v) = 4π (m / [2π*k*_BT])^3/2 v² exp(−*mv*² / [2*k*_BT])

The competition between the quadratic phase-space factor v² (which vanishes as v → 0) and the decaying Boltzmann exponential factor exp(−*mv*² / 2*k*_BT) produces an asymmetric distribution characterized by three ordered characteristic velocities: the most probable speed v_mp = √(2*k*_BT / m), the mean speed ⟨*v*⟩ = √(8*k*_BT / π*m*) ≈ 1.128 v_mp, and the root-mean-square speed v_rms = √(3*k*_BT / m) ≈ 1.225 v_mp.

Two-State Biophysics: Protein Folding and Ion Channel Gating

In molecular biophysics, macromolecules at constant temperature T and pressure P populate conformational macrostates (such as the folded native state N and unfolded denatured ensemble U in protein folding) according to their Gibbs free energy G_I = H_I − TS_I, where S_I accounts for the conformational degeneracy of macrostate I and surrounding H₂O solvent molecules. The equilibrium ratio of probabilities between two conformational states is governed by the Boltzmann relation:

K_eq = P_N / P_U = exp(−Δ*G*° / [RT]) = exp(−[Δ*H*° − *T*Δ*S*°] / [RT])

At physiological temperature (T = 310 K, where RT = 2.58 kJ mol⁻¹ = 0.616 kcal mol⁻¹), a modest folding free energy difference of Δ*G*° = −25.8 kJ mol⁻¹ (−10 RT) yields an equilibrium ratio P_N / P_U = e¹⁰ ≈ 22,026, ensuring that 99.995% of protein molecules reside in the functional native fold despite continual thermal agitation.

Similarly, in cellular electrophysiology, a voltage-gated Na⁺ or K⁺ ion channel transitions between a closed conformation (C) and an open pore conformation (O). When the transmembrane electrical potential changes to V_m, a gating charge z_ge moves across the membrane electric field, shifting the free energy difference between the open and closed states to Δ*G*(V_m) = z_ge(V_1/2 − V_m), where V_1/2 is the half-activation voltage. Because the channel must be either open or closed (P_open + P_closed = 1), the Boltzmann ratio P_open / P_closed = exp(−Δ*G* / k_BT) yields the sigmoidal Hodgkin–Huxley activation curve:

P_open(V_m) = 1 / [1 + exp(z_ge(V_1/2 − V_m) / [k_BT])]

How does the Boltzmann distribution relate to the Bose–Einstein and Fermi–Dirac distributions?

The Boltzmann distribution is the classical high-temperature or low-density limit of quantum statistical mechanics, valid when the mean thermal de Broglie wavelength λ_th = h / √(2π*mk*_BT) is much smaller than the average interparticle spacing d = (V/N)^1/3, so that the average occupancy of any single-particle quantum state is much less than 1. At high densities or cryogenic temperatures where wavefunctions overlap, indistinguishable particles with half-integer spin (fermions, such as electrons) obey the Pauli exclusion principle and follow the Fermi–Dirac distribution n_i = [exp((E_i − μ)/k_BT) + 1]⁻¹, whereas integer-spin particles (bosons, such as photons in the cosmic microwave background or ⁴He atoms) follow the Bose–Einstein distribution n_i = [exp((E_i − μ)/k_BT) − 1]⁻¹.

Can the Boltzmann distribution produce a higher population in a higher-energy state?

For individual non-degenerate microstates at positive absolute temperature (T > 0), a higher-energy state E₂ > E₁ always has a strictly lower probability (p₂ < p₁) in thermal equilibrium. However, if an upper energy level has a much larger degeneracy g₂ ≫ g₁ (such as rotational J levels in diatomic molecules where g_J = 2*J* + 1), the level population P(E_J) ∝ (2*J* + 1) exp(−*E*_J/k_BT) peaks at J > 0. True microstate population inversion (p₂ > p₁ for E₂ > E₁), as required for laser amplification or nuclear spin systems, requires driving the system out of thermal equilibrium via optical pumping (formally described by a negative effective spin temperature T < 0 in bounded-energy systems).

1. Boltzmann, L. (1877). "Über die Beziehung zwischen dem zweiten Hauptsatze der mechanischen Wärmetheorie und der Wahrscheinlichkeitsrechnung respektive den Sätzen über das Wärmegleichgewicht." Wiener Berichte, 76, 373–435. doi:10.1017/CBO9781139381437.011
2. Gibbs, J. W. (1902). Elementary Principles in Statistical Mechanics: Developed with Especial Reference to the Rational Foundation of Thermodynamics. Charles Scribner's Sons. doi:10.1017/CBO9780511686948
3. Jaynes, E. T. (1957). "Information Theory and Statistical Mechanics." Physical Review, 106(4), 620–630. doi:10.1103/PhysRev.106.620
4. Dill, K. A., & Bromberg, S. (2010). Molecular Driving Forces: Statistical Thermodynamics in Biology, Chemistry, Physics, and Nanoscience (2nd ed., Chapter 6: "Boltzmann's Distribution Law"). Garland Science. https://doi.org/10.1201/9780429258817