The Maxwell–Boltzmann distribution is the probability distribution of particle velocities in a classical system at thermal equilibrium. Its best-known form gives the distribution of molecular speeds in an ideal gas: at a given temperature, particles have a continuous range of speeds rather than a single common speed. The distribution connects molecular motion with measurable properties such as pressure, internal energy, diffusion, and reaction rates, making it a central result of statistical mechanics and the kinetic theory of gases.
The name encompasses closely related distributions for the three-dimensional velocity vector, individual velocity components, speed, and translational kinetic energy. These must be distinguished because they have different mathematical forms and different most probable values. The distribution applies when particle motion is nonrelativistic and quantum exchange effects are negligible. Although usually introduced for dilute gases, its equilibrium momentum form also applies to many interacting classical systems.
Physical meaning and conditions of applicability
A gas in thermodynamic equilibrium has stationary macroscopic properties, but its molecules remain in motion. Collisions continually change individual velocities while leaving the equilibrium distribution unchanged. The Maxwell–Boltzmann distribution therefore describes a population statistically; it does not assign each molecule a permanent speed or prescribe the detailed trajectory of any particular particle.
Velocity and speed are distinct quantities. Velocity, , includes direction, whereas speed, , is nonnegative. In a gas at rest, velocities pointing in opposite directions are equally probable, so the mean velocity vanishes. The mean speed does not vanish. A gas can consequently have zero net flow while its particles move at hundreds of metres per second.
The relevant frame is the local rest frame of the material. If a gas has a uniform bulk velocity , its thermal motion is described by the peculiar velocity
A Maxwellian distribution in becomes a shifted Maxwellian in laboratory velocity. This distinction is essential in flowing gases and plasmas, where directed motion and random thermal motion can contribute very differently to the total kinetic energy. Temperature characterizes the latter, not the kinetic energy associated with the motion of the entire sample.
The classical assumptions concern both dynamics and statistics. Particle energies must be well approximated by , as in classical mechanics, rather than by a relativistic expression. Quantum occupation effects must also be small. Neither condition requires every particle to be distinguishable in principle; rather, the observable consequences of quantum exchange must be negligible in the regime being described.
An ideal gas is the simplest setting, but the velocity distribution has a wider domain. For a classical equilibrium system whose Hamiltonian separates into ordinary quadratic kinetic energy and a position-dependent interaction energy, the momentum distribution is Maxwellian even when spatial correlations are strong. Intermolecular forces can change the equation of state and collision dynamics without changing this equilibrium momentum marginal. Conversely, a dilute gas need not be Maxwellian if it is driven, rapidly expanding, or insufficiently collisional. Low density alone does not establish equilibrium.
Velocity, speed, and energy distributions
For particles of mass at temperature , the normalized three-dimensional velocity density in the rest frame is
where is the Boltzmann constant. The probability of finding a particle in a small velocity-space volume is , and integration over all velocities gives unity. A number-density distribution is instead written , where is the particle number per unit physical volume.
Each Cartesian component follows a normal distribution:
The three components are independent in this equilibrium distribution. Their means are zero and their variances are . The component density has its maximum at zero; this is fully consistent with a nonzero most probable speed because a speed interval includes many different velocity directions.
For an isotropic distribution, the velocity-space shell between radii and has volume . The speed density is therefore
The factor is geometric. It counts the increasing volume of velocity space available at larger speeds, while the exponential suppresses high kinetic energies. Their competition produces a peak at a positive speed. A probability density is not itself the probability of one exact speed: probabilities are obtained by integrating over finite intervals.
Changing variables to translational energy, , gives
This energy distribution is a gamma distribution with shape parameter . At a common temperature it is independent of particle mass, although the corresponding speed distribution is not. Heavy particles move more slowly on average, but possess the same mean translational energy as light particles. The prefactor also shows why an energy distribution cannot generally be identified with the exponential Boltzmann factor alone: the number of accessible states must be included.
Derivation from equilibrium statistical mechanics
The most direct equilibrium derivation uses the Boltzmann distribution. In a system exchanging energy with a reservoir at temperature , a microstate of energy has a probability proportional to , with . For a free particle, , so the momentum density is a Gaussian. Substitution of , followed by normalization, gives the Maxwellian velocity density.
For interacting particles with ordinary classical kinetic energy, the Hamiltonian can be written
The canonical weight then factorizes:
Integrating over positions removes the configurational factor from the normalized momentum marginal. This explains why a Maxwellian momentum distribution can hold in an interacting liquid as well as in an ideal gas. The conclusion depends on the form of the Hamiltonian; it is not a statement that interactions are unimportant for every property.
For a single free particle in volume , the translational partition function is
where is the Planck constant and
is the thermal de Broglie wavelength in a standard convention. The phase-space normalization involving is necessary for counting states consistently, but it cancels from the normalized classical velocity distribution. For an indistinguishable ideal gas, the classical many-particle partition function includes the factor .
A second derivation maximizes entropy subject to fixed normalization and mean kinetic energy. Variation of the continuous entropy functional gives a density proportional to ; the energy constraint fixes . This method makes explicit that the equilibrium distribution is selected by constraints, rather than by an assumption that all speeds are equally likely.
Maxwell’s original reasoning instead combined rotational symmetry with statistical independence of orthogonal velocity components. The resulting functional equation admits Gaussian component densities. Isotropy alone is insufficient: many non-Gaussian distributions are rotationally symmetric. Likewise, the central limit theorem offers useful intuition about accumulated random influences but is not required for the equilibrium statistical-mechanical derivation. The constraints and assumptions determine which argument is valid.
Characteristic speeds and thermodynamic quantities
Three characteristic speeds appear frequently:
They are, respectively, the most probable speed, the mean speed, and the root-mean-square speed. Their ordering is
because the speed distribution has a high-speed tail. The most probable speed is the mode of the speed density, not of the velocity-vector density or the energy density. In particular, the most probable translational energy is , whereas the energy evaluated at the most probable speed is . Modes do not generally remain at equivalent locations after a nonlinear change of variables.
All speed moments can be written as
where is the gamma function. The second moment gives
Each quadratic translational degree of freedom contributes , an instance of the equipartition theorem. Molecular rotation and vibration require separate treatment; they are not included in the three-dimensional translational speed distribution.
For nitrogen molecules at , using a molar mass of approximately , the characteristic speeds are about , , and . The same formulas can be expressed using in place of , where is the molar gas constant and is molar mass. At fixed temperature, every characteristic speed scales as ; at fixed mass, it scales as .
The mechanical interpretation of pressure follows from momentum transfer to container walls:
Isotropy gives . For a monatomic ideal gas, the translational internal energy is , and the constant-volume molar heat capacity is . In an interacting fluid, the same Maxwellian kinetic contribution can coexist with an additional configurational contribution to pressure and internal energy; the ideal-gas equation of state does not follow from the velocity distribution alone.
Collisions, relaxation, and transport
Elastic collisions redistribute velocity while conserving total momentum and energy. For a suitable dilute gas, the evolution of its one-particle distribution is described by the Boltzmann equation:
where is an external force and is the collision operator. The notation separates streaming through physical space, acceleration through velocity space, and collisional redistribution.
For a spatially uniform equilibrium Maxwellian, the elastic collision operator vanishes. In a binary collision, conservation of total kinetic energy ensures that the product of the two incoming Maxwellian factors equals the corresponding product for the outgoing particles. Gain and loss terms consequently balance. Under the kinetic equation’s assumptions, the H-theorem describes a monotonic tendency toward equilibrium. Its statistical premise includes the factorization of precollision pair distributions, commonly called molecular chaos; this premise is important when relating kinetic irreversibility to reversible microscopic mechanics.
In a hard-sphere model with molecular diameter , the mean free path is
A characteristic collision time is of order . Actual relaxation times depend on the interaction potential and on the quantity being relaxed. Momentum isotropization, energy exchange between species, and equilibration of internal molecular modes can occur on different timescales. A single “collision time” therefore does not describe every approach to equilibrium.
Transport arises when the distribution is close to, but not exactly, a local Maxwellian. Particles carry momentum across velocity gradients, producing viscosity; they carry energy across temperature gradients, producing thermal conductivity; and they redistribute species across concentration gradients, producing diffusion. If a distribution were exactly isotropic in each local rest frame, its peculiar-velocity third moment—and hence its conductive heat flux—would vanish. Small anisotropic corrections are responsible for these dissipative fluxes.
The ratio , the Knudsen number, compares mean free path with a macroscopic variation length . Small values often support a local-equilibrium expansion and continuum fluid equations. At larger values, boundary effects and nonlocal particle motion become important. In mixtures, collisions tend to establish a common bulk velocity and temperature, but not identical speed distributions: each species retains the mass dependence of its own Maxwellian. Strong shocks, rarefied expansions, and weakly collisional plasmas can exhibit substantial departures from this local-equilibrium picture.
Historical development and experimental foundations
James Clerk Maxwell introduced the velocity distribution in 1860 in “Illustrations of the Dynamical Theory of Gases.” Earlier kinetic models often represented molecular motion using a characteristic or mean speed. Maxwell’s statistical treatment instead assigned probabilities to different velocities, permitting measurable gas properties to be calculated without following the detailed paths of an enormous number of particles.
The result belonged to a broader nineteenth-century effort to explain thermodynamics through mechanics. Molecular motion provided an interpretation of temperature, while impacts on boundaries explained pressure. Maxwell also used kinetic reasoning to investigate transport. A notable prediction was that, over the dilute-gas regime where the model applies, gas viscosity is approximately independent of density: decreasing density reduces the number of momentum carriers but increases the distance they travel between collisions. This cancellation was a distinctive test of kinetic theory, though it did not directly map out the entire speed distribution.
Ludwig Boltzmann extended the statistical theory in a series of works beginning in the 1860s. In 1872 he formulated the kinetic equation and H-theorem, connecting collisional dynamics with a tendency toward the equilibrium distribution. His 1877 probabilistic treatment of entropy developed the combinatorial interpretation of macrostates. The combined name “Maxwell–Boltzmann” reflects this broader synthesis, although Maxwell’s velocity law, the Boltzmann energy factor, and classical state-occupation statistics are related but distinct statements.
Direct velocity measurements became possible through molecular-beam techniques. In 1920, Otto Stern reported a molecular-beam determination of molecular thermal speeds using silver atoms and rotating apparatus. Such experiments converted microscopic motion into spatial displacements or arrival-time differences that could be measured. Their interpretation required attention to the way particles were extracted from the source, because a beam does not automatically sample the equilibrium population without bias.
Later methods included time-of-flight measurements and spectroscopic determinations of thermal velocity components. These approaches test different aspects of the theory: a beam measurement may probe a flux-weighted speed distribution, whereas a thermal spectral line can reveal a one-dimensional velocity distribution. Agreement among such measurements supports the statistical description more broadly than a single measurement of mean speed would.
The historical importance of the distribution is consequently both conceptual and empirical. It demonstrated how a precise mathematical law could describe molecular disorder, and it helped establish probability as a fundamental working language of physical theory. The development of quantum statistics in the twentieth century did not discard the classical distribution; it identified the conditions under which it emerges as an accurate limiting description.
Measurement, effusion, and sampling effects
A measured particle distribution depends on how particles enter the instrument. Particles counted at randomly chosen instants within a volume sample the population distribution. Particles crossing a surface sample a flux distribution: those with a larger velocity component toward the surface cross it more frequently. Confusing these two sampling procedures produces systematic discrepancies even when the source gas is perfectly Maxwellian.
For an equilibrium gas next to a small planar aperture, the number flux crossing the aperture is
The aperture must operate in the effusive regime: it should be small relative to the mean free path, so that passage through the opening does not involve substantial interparticle collisions. A large opening or a sufficiently high source pressure can instead produce a collisional expansion with a markedly different velocity distribution.
For particles emerging effusively, the normalized speed density is
Its most probable speed is , greater than the most probable speed in the source volume. The angular flux from an ideal planar aperture follows a cosine law relative to the surface normal. Subsequent collimation and detector acceptance can further alter the observed distribution, particularly if angular selection and speed sensitivity are coupled.
Time-of-flight measurements introduce another change of variables. For a fixed flight distance , arrival time is , so a speed density becomes
The Jacobian is essential. An arrival-time histogram cannot be interpreted as a speed histogram merely by relabelling its horizontal axis. The distribution used in this transformation must also be the one appropriate to the source and detection process—for example, a beam-flux distribution rather than an equilibrium volume distribution.
Thermal Doppler broadening provides a complementary measurement. A spectral line samples the velocity component along the observation direction. In the nonrelativistic limit, a transition of rest frequency has thermal frequency standard deviation
and Gaussian full width at half maximum . Here is the speed of light. In practical spectroscopy, natural linewidth, collisions, unresolved flows, and instrumental response can add other broadening mechanisms. Thermal interpretation therefore rests on a model of the complete line profile, not width alone.
Chemical, atmospheric, and computational applications
Collision rates depend on relative velocities rather than on the speed of one arbitrarily selected particle. For two independent Maxwellian species at the same temperature, with masses and , the relative velocity is itself Maxwellian with the reduced mass
Its mean magnitude is
For identical particles this is times the mean single-particle speed. This distinction matters in estimating collision frequencies and mean free paths.
In chemical kinetics, a bimolecular rate coefficient can be expressed as a thermal average of a reactive cross section :
The energy is the relative translational energy. This expression shows how a distribution of collision energies combines with microscopic reaction dynamics. An activation barrier can make the high-energy tail disproportionately important, but the Maxwellian factor alone does not determine the rate. Molecular orientation, internal excitation, and quantum tunnelling may also affect the cross section.
Atmospheric escape provides another application of the tail. A particle at radius from a spherical body of mass requires speed to escape gravitationally in the simplest model. The dimensionless escape parameter is
For an approximately Maxwellian population at an exobase, the idealized Jeans escape flux is
Its exponential sensitivity helps explain why light species escape more readily. The model assumes that the source population remains close to Maxwellian and does not describe all atmospheric loss mechanisms, such as hydrodynamic outflow or nonthermal energization.
In nuclear fusion, thermal reaction rates similarly average cross sections over relative energies. For charged nuclei, the decreasing Maxwellian factor competes with the increasing probability of penetrating the Coulomb barrier. The effective reaction-energy window therefore need not lie near the most probable thermal energy. In laboratory and astrophysical plasmas, departures from a Maxwellian can change such averages even when the mean energy is unchanged.
Computationally, the distribution supplies initial velocities for molecular dynamics and particle-based gas simulations. Independent Gaussian components are drawn with variance ; any unwanted total momentum can then be removed. Removing center-of-mass motion reduces the number of independent kinetic degrees of freedom. Exact rescaling to a prescribed instantaneous kinetic energy also changes finite-system statistics, so initialization, temperature estimation, and thermostat design must be treated consistently.
Quantum, relativistic, and nonequilibrium limits
Classical Maxwell–Boltzmann statistics emerge when quantum states are sparsely occupied. For a nonrelativistic ideal gas, a useful degeneracy parameter is
where is the internal degeneracy. Values much smaller than unity identify a regime in which quantum exchange corrections are generally small. Increasing density or lowering temperature causes thermal wave packets to overlap more strongly, making particle identity relevant to state occupation.
For a single-particle state of energy , Bose–Einstein statistics and Fermi–Dirac statistics give, respectively,
Here is the chemical potential. When the exponential is much larger than unity, both expressions reduce to
the classical occupation law. Outside that limit, bosonic enhancement and the Pauli exclusion principle produce distributions that cannot be represented by the classical Maxwellian. Degenerate electrons in metals and ultracold atomic gases are important examples.
The relativistic limitation is different. The classical distribution extends to arbitrarily large speeds, including speeds above . At ordinary molecular temperatures, the formal weight of this unphysical tail is negligible. When becomes comparable to , however, the quadratic energy approximation fails. For a dilute, nondegenerate relativistic gas, the Maxwell–Jüttner distribution replaces the classical law, using the relativistic energy in the equilibrium momentum weight. Transforming that momentum distribution into a velocity distribution requires the appropriate relativistic Jacobian.
Nonequilibrium departures do not require either quantum degeneracy or relativistic motion. Particle beams, electrical acceleration, shocks, and weakly collisional space plasmas can produce drifting populations, anisotropic distributions, or enhanced high-energy tails. A bi-Maxwellian model assigns different component temperatures parallel and perpendicular to a preferred direction, often a magnetic field. Such a model may be useful without describing full thermodynamic equilibrium.
Finally, a measured second moment is insufficient to establish Maxwellian behavior. A non-Gaussian population can have exactly the same mean kinetic energy as a Maxwellian while having different collision rates, escape fluxes, or spectral wings. Tests therefore examine distribution shape, component anisotropy, and higher moments as well as an inferred temperature. Even in classical equilibrium, a finite isolated system with exactly fixed total energy and momentum need not have an exactly Maxwellian one-particle marginal; the familiar form emerges in the appropriate large-system limit or canonical description.