The Boltzmann distribution is a probability distribution that describes how a physical system occupies its possible microscopic states at thermal equilibrium. Each state receives a statistical weight that decreases exponentially with its energy at positive temperature. A foundational result of statistical mechanics, it connects microscopic states with macroscopic thermal behavior and supplies the equilibrium probabilities for the canonical ensemble. (damtp.cam.ac.uk)
Mathematical definition
For discrete states indexed by , with energies , the probability of state is
Here is the Boltzmann constant, is absolute temperature, and is the partition function. The numerator, called the Boltzmann factor, is an exponential function of energy. Division by ensures that the probabilities sum to one; the distribution requires this normalization to be finite. (damtp.cam.ac.uk)
For two individual states,
Thus only energy differences determine their relative probabilities. Adding the same constant to every energy leaves the distribution unchanged. The quantity provides the relevant thermal energy scale: an energy separation much larger than this scale strongly suppresses the higher state. These consequences follow directly from the canonical probability formula. (damtp.cam.ac.uk)
Physical origin and assumptions
The standard derivation considers a system exchanging energy with a much larger thermal reservoir while its volume and particle number remain fixed. The combined system and reservoir are isolated, with total energy . Assuming equally probable accessible microstates of the combined object, the probability of system state is proportional to the reservoir’s number of states at energy . (interactivetextbooks.tudelft.nl)
Writing this multiplicity in terms of reservoir entropy gives
For a sufficiently large reservoir, its temperature changes negligibly as energy passes between the two parts. Expanding its entropy to first order,
produces the Boltzmann factor after normalization. The derivation therefore depends on equilibrium and on a reservoir whose temperature is effectively fixed. It describes equilibrium occupancy, rather than the rate at which equilibrium is reached. (interactivetextbooks.tudelft.nl)
States, energy levels, and degeneracy
An individual microstate must be distinguished from an energy level containing several states. If level has energy and degeneracy , its total probability is
The degeneracy counts independent states sharing that energy. A highly degenerate level can consequently have a larger total population than a less degenerate lower level, even though each of its individual states has a smaller probability. This distinction matters when interpreting thermal populations in spectroscopy. (physchem.no)
As an illustrative consequence, two nondegenerate levels with energies and have upper-level probability
It approaches zero as and one-half as . These limits describe this finite two-level model, not a general uniform distribution over an unlimited number of states. (physchem.no)
Thermodynamic quantities and fluctuations
The partition function connects state probabilities with thermodynamics. The expected value of energy, identified with the system’s internal energy, is
Its Helmholtz free energy is
Thus is more than a normalization constant: its dependence on temperature and other system parameters determines equilibrium thermodynamic properties. Computing it can nevertheless be difficult when the number of configurations is large or interactions are complicated. (interactivetextbooks.tudelft.nl)
Energy fluctuates because the system exchanges heat with its reservoir. For temperature-independent state energies at fixed volume and particle number,
where is the constant-volume heat capacity. This relation links microscopic energy variance to a measurable macroscopic response. (interactivetextbooks.tudelft.nl)
Classical systems
In classical mechanics, states form a continuous phase space of positions and momenta. The corresponding probability density is proportional to , where is the system’s Hamiltonian, and sums over states become integrals with an appropriate phase-space measure. (damtp.cam.ac.uk)
For a nonrelativistic ideal gas, the kinetic energy produces Gaussian velocity components. Accounting for the number of velocity vectors having a given speed yields the three-dimensional Maxwell–Boltzmann speed distribution,
The factor shows why the speed distribution is not simply the Boltzmann factor evaluated at kinetic energy. (damtp.cam.ac.uk)
Relation to quantum statistics
The canonical Boltzmann distribution also applies in quantum mechanics: its states can be the energy eigenstates of an entire many-particle system. It is therefore not restricted to classical particles. (damtp.cam.ac.uk)
However, probabilities for complete system states are different from mean occupations of single-particle states. For ideal quantum gases, those occupations follow Bose–Einstein statistics or Fermi–Dirac statistics. Both reduce to the classical exponential form when occupation numbers are small:
where is the chemical potential. Quantum occupation laws therefore do not invalidate canonical Boltzmann weights; they arise from applying equilibrium statistical mechanics with the appropriate many-particle state counting. (damtp.cam.ac.uk)