Boltzmann distribution

In classical statistical physics, the Boltzmann distribution expresses the relative probability that a subsystem of a physical system has a certain energy. The subsystem must part of a physical system that is in thermal equilibrium, that is, the system must have a well-defined (absolute) temperature. For instance, a subsystem can be a single molecule in, say, one mole of an ideal gas. Then the Boltzmann distribution applies to the energies of the individual gas molecules, provided the ideal gas is in thermal equilibrium.

The Boltzmann distribution (also known as the Maxwell-Boltzmann distribution) was proposed in 1859 by the Scotsman James Clerk Maxwell for the statistical distribution of the kinetic energies of ideal gas molecules. Consider an ideal gas of absolute temperature T. Let n₁ be the number of molecules with kinetic energy E₁ and n₂ be the number with kinetic energy E₂, then according to the Maxwell-Boltzmann distribution law the relative probability is the ratio,

${\frac {n_{1}}{n_{2}}}={\frac {e^{-E_{1}/(kT)}}{e^{-E_{2}/(kT)}}},\qquad \qquad \qquad \qquad (1)$

where k is the Boltzmann constant. Most noticeable in this expression are (i) the energy in an exponential, (ii) the inverse temperature in the exponent, and (iii) the appearance of the natural constant k. Note that an argument of an exponential must be dimensionless and that accordingly kT has the dimension energy.

As discovered in 1871 by the Austrian Ludwig Boltzmann, the molecular energies in equation (1) may, in addition to the translational energy considered by Maxwell, contain rotational and vibrational energies of the molecules. Also interactions with an external field may be included. If a system, as for instance a column of air of constant temperature (zero lapse rate), is in the gravitational field of the Earth, each molecular energy contains the additional term mgh, where m is the molecular mass, g the gravitational acceleration and h the height of the molecule above the surface of the Earth. Thus, the ratio of the number of molecules at height h₁ and h₂ is,

${\frac {n(h_{1})}{n(h_{2})}}={\frac {e^{-(E_{1}+mgh_{1})/(kT)}}{e^{-(E_{2}+mgh_{2})/(kT)}}}\approx \exp {\Big [}-{\frac {mg(h_{1}-h_{2})}{kT}}{\Big ]},$

where we made the assumption that the difference in molecular kinetic energies at the two heights is small in comparison to the difference in potential energies:

E₂ − E₁ << mg(h₂−h₁),

so that the difference in kinetic energies may be omitted.

Generalization

A few years after Boltzmann, the American Josiah Willard Gibbs gave a further formalization and generalization (ca. 1877). Gibbs introduced what he called an ensemble, a "supersystem" consisting of a statistically large number of identical systems. For instance, the systems may be identical vessels containing the same number of the same real gas molecules at the same temperature and pressure. Further Gibbs assumed that the ensemble, just like a system of gas molecules, is in thermal equilibrium, i.e., the systems are in thermal contact so that they can exchange heat. In the example that the systems are gas-filled vessels, this is achieved by requiring that the vessels are in mutual contact through heat-conducting walls.

Gibbs assumed that the Maxwell-Boltzmann law, equation (1), holds for the energies of the systems in the ensemble. This generalization, applied to the example of an ensemble consisting of gas-filled vessels, means that the energy of a gas molecule in equation (1) is replaced by the total energy of the molecules in a vessel; a "one-molecule" energy is promoted to a "one-vessel" energy. Because of molecular interactions (that are absent in an ideal gas), a one-vessel energy cannot be written as a sum of one-molecule energies. This is the main reason for Gibbs' generalization from a single vessel of gas to an ensemble of vessels, or, in more general terms, the generalization from an ideal gas to an ensemble of (rather arbitrary) systems. Just as the molecules in an ideal gas, the systems in the ensemble do not interact other than by exchanging heat.

The absolute probability for a system to have total energy ε_j can be obtained from equation (1) by normalizing. Let us assume, for convenience sake, that the one-system energies are discrete (as they often are in quantum mechanics) running from ε₀ to ε_∞ , then

$N_{j}e^{-\epsilon _{i}/(kT)}=N_{i}e^{-\epsilon _{j}/(kT)}\;\Longrightarrow \;N_{j}\sum _{i=0}^{\infty }e^{-\epsilon _{i}/(kT)}=e^{-\epsilon _{j}/(kT)}\sum _{i=0}^{\infty }N_{i}$

which gives

${\frac {N_{j}}{N}}={\frac {e^{-\epsilon _{j}/(kT)}}{Q}}\quad {\hbox{with}}\quad Q\equiv \sum _{i=0}^{\infty }e^{-\epsilon _{i}/(kT)}\quad {\hbox{and}}\quad N=\sum _{i=0}^{\infty }N_{i},$

where N_j is the number of systems that have energy ε_j and N is the total number of systems in the ensemble (N must be very large—go to infinity—for statistics to apply). The ratio is a probability written as

${\frac {N_{j}}{N}}\equiv {\mathcal {P}}(\epsilon _{j})={\frac {e^{-\epsilon _{j}/(kT)}}{Q}}.$

The quantity Q is known as the partition function of the system. It is a sum over all energies that the system can have. In the older literature Q is called Zustandssumme, which is German for "sum over states", and is often denoted by Z. When we consider again the example of a vessel with interacting molecules at certain pressure and temperature, the sum in the partition function is over the possible total energies ε_i of the molecules in the vessel.

In classical statistical physics, where energies are not discrete, the partition function of a system of n molecules is not a sum over the possible energies of the system, but an integral over the 6ⁿ-dimensional phase space (space of momenta and positions). In the framework of the "old quantum theory" it was discovered in 1912 that the classical partition function of n molecules must multiplied by a quantum factor. The classical partition function is,

$Q_{\mathrm {class} }={\frac {1}{n!h^{3n}}}\int e^{-\epsilon (p,q)/(kT)}\mathrm {d} \mathbf {q} _{1}\cdots \mathrm {d} \mathbf {q} _{n}\,\mathrm {d} \mathbf {p} _{1}\cdots \mathrm {d} \mathbf {p} _{n},$

where h is Planck's constant and n! = 1 × 2× ... × n (the factorial of n).

As a final remark: in quantum statistical thermodynamics the Boltzmann distribution appears as the high-temperature limit of both Bose-Einstein statistics (valid for bosons) and Fermi-Dirac statistics (valid for fermions).