Derivation
The Maxwell–Boltzmann (MB) distribution
describes the velocities
or the kinetic energy
of the particles at thermal equilibrium, far from the limit of the speed of light, i.e.:
![{\displaystyle {\begin{aligned}f_{\text{MB}}(\mathbf {p} ;v_{\text{th}})\,\mathrm {d} ^{n}\mathbf {p} &=\left(\pi m^{2}v_{\text{th}}^{2}\right)^{-n/2}e^{-{\frac {\mathbf {p} ^{2}/2m}{k_{\text{B}}T}}}\,\mathrm {d} ^{n}\mathbf {p} ,&v_{\text{th}}\equiv {\sqrt {2{k_{\text{B}}T}/{m}}},\\[1ex]f_{\text{MB}}(\varepsilon ;T)\,\mathrm {d} \varepsilon &={\frac {\left(k_{\text{B}}T\right)^{-n/2}}{\Gamma {\left(n/2\right)}}}e^{-\varepsilon /{k_{\text{B}}T}}\varepsilon ^{{\frac {1}{2}}n-1}\,\mathrm {d} \varepsilon \end{aligned}}}](./_assets_/eb734a37dd21ce173a46342d1cc64c92/e12145c78a9a90f330319c4ec50d73c9c2defe8a.svg) | | 1 |
where the low-energy limit (non-relativistic) can be expressed as
, or equivalently,
. Here,
is the expression of temperature as a speed, called thermal speed, and n denotes the kinetic degrees of freedom of each particle. (Note that the temperature is defined in the fluid's rest frame, where the bulk speed
is zero. In the non-relativistic case, this can be shown by using
.
The relativistic generalization of Eq. (1), that is, the Maxwell–Jüttner (MJ) distribution, is given by:
 | | 2 |
where
,
, and
. (Note that the inverse of the unitless temperature
is the relativistic coldness
.[6]) This distribution (Eq. 2) can be derived as follows. According to the relativistic formalism for the particle momentum and energy, one has

While the kinetic energy is given by
. The Boltzmann distribution of a Hamiltonian is
. In the absence of a potential energy,
is simply given by the particle energy
, thus:

(Note that
is the sum of the kinetic
and rest energy
,
). Then, when one includes the n-dimensional density of states:

So that:
![{\displaystyle {\begin{aligned}\int f_{\text{MJ}}(\mathbf {p} )\,\mathrm {d} p_{1}\cdots \mathrm {d} p_{n}&\propto \int e^{-{\frac {E(\mathbf {p} )}{k_{\text{B}}T}}}\,\mathrm {d} p_{1}\cdots \mathrm {d} p_{n}\\[1ex]&=\int e^{-{\frac {E(\gamma \Omega _{n})}{k_{\text{B}}T}}}\,\mathrm {d} \Omega _{n}\,p^{n-1}\,\mathrm {d} p\\[1ex]&=\int _{\Omega _{n}}e^{-{\frac {E(\gamma \Omega _{n})}{k_{\text{B}}T}}}\,\left(p(\gamma )^{n-1}{\frac {\mathrm {d} p(\gamma )}{\mathrm {d} \gamma }}\right)\mathrm {d} \Omega _{n}\,\mathrm {d} \gamma \end{aligned}}}](./_assets_/eb734a37dd21ce173a46342d1cc64c92/82258893125a1d79598d0aee1c9f2ee5bba54bd3.svg)
Where
denotes the n-dimensional solid angle. For isotropic distributions, one has

or

Then,
so that:

Or:
 | | 3 |
Now, because
. Then, one normalises the distribution Eq. (3). One sets

And the angular integration:

Where
is the surface of the unit n-dimensional sphere. Then, using the identity
one has:
;\theta )\,\mathrm {d} p_{1}\cdots \mathrm {d} p_{n}={\frac {N}{2}}B_{n}\left(mc\right)^{n}\,e^{-\gamma /\theta }\left(\gamma ^{2}-1\right)^{{\frac {n}{2}}-1}\,\mathrm {d} (\gamma ^{2}-1).}
and
;\theta )\,\mathrm {d} p_{1}\cdots \mathrm {d} p_{n}\\[1ex]&={\frac {N}{2}}B_{n}\left(mc\right)^{n}\,\int _{1}^{\infty }e^{-{\frac {\gamma }{\theta }}}\left(\gamma ^{2}-1\right)^{{\frac {n}{2}}-1}\mathrm {d} (\gamma ^{2}-1)\\[1ex]&={\frac {N}{2}}B_{n}\left({\frac {n}{2}}\right)^{-1}\left(mc\right)^{n}\theta ^{-1}\,\int _{1}^{\infty }e^{\frac {\gamma }{\theta }}\left(\gamma ^{2}-1\right)^{\frac {n}{2}}\mathrm {d} \gamma \\[1ex]&={\frac {N}{2}}B_{n}\left({\frac {n}{2}}\right)^{-1}\left(mc\right)^{n}\theta ^{-1}\,I_{n},\end{aligned}}}
![{\displaystyle {\begin{aligned}1&=\int _{-\infty }^{\infty }f_{\text{MJ}}(\mathbf {p} ;\theta )\,\mathrm {d} p_{1}\cdots \mathrm {d} p_{n}\\[1ex]&={\frac {N}{2}}B_{n}\left(mc\right)^{n}\,\int _{1}^{\infty }e^{-{\frac {\gamma }{\theta }}}\left(\gamma ^{2}-1\right)^{{\frac {n}{2}}-1}\mathrm {d} (\gamma ^{2}-1)\\[1ex]&={\frac {N}{2}}B_{n}\left({\frac {n}{2}}\right)^{-1}\left(mc\right)^{n}\theta ^{-1}\,\int _{1}^{\infty }e^{\frac {\gamma }{\theta }}\left(\gamma ^{2}-1\right)^{\frac {n}{2}}\mathrm {d} \gamma \\[1ex]&={\frac {N}{2}}B_{n}\left({\frac {n}{2}}\right)^{-1}\left(mc\right)^{n}\theta ^{-1}\,I_{n},\end{aligned}}}](./_assets_/eb734a37dd21ce173a46342d1cc64c92/82df0ebf0e8fb9dbd501e1c6f44ef624f3972c3b.svg)
Where one has defined the integral:

The Macdonald function (Modified Bessel function of the second kind) (Abramowitz and Stegun, 1972, p.376) is defined by:

So that, by setting
one obtains:

Hence,
![{\displaystyle {\begin{aligned}N^{-1}&={\frac {\pi ^{\frac {n}{2}}}{\Gamma {\left({\frac {n}{2}}\right)}}}\left({\frac {n}{2}}\right)^{-1}\Gamma {\left({\frac {n}{2}}+1\right)}\pi ^{-{\frac {1}{2}}}\operatorname {K} _{\frac {n+1}{2}}{\!\!\left({\frac {1}{\theta }}\right)}\,\left(mc\right)^{n}\left(2\theta \right)^{\frac {n+1}{2}}\\[1ex]&=\pi ^{\frac {n-1}{2}}2^{-{\frac {n+1}{2}}}\left(mc\right)^{n}\,\theta ^{\frac {n-1}{2}}\operatorname {K} _{\frac {n+1}{2}}{\!\!\left({\frac {1}{\theta }}\right)},\end{aligned}}}](./_assets_/eb734a37dd21ce173a46342d1cc64c92/35f48bd705ed007e956890a0c8fa90b77fc4a728.svg)
Or

The inverse of the normalization constant gives the partition function
:

Therefore, the normalized distribution is:

Or one may derive the normalised distribution in terms of:
;\theta )\,\mathrm {d} \gamma ={\frac {\pi ^{\frac {1}{2}}2^{\frac {1-n}{2}}}{\Gamma {\left({\frac {n}{2}}\right)}}}\theta ^{\frac {1-n}{2}}{\frac {e^{-\gamma /\theta }}{\operatorname {K} _{\frac {n+1}{2}}{\!\!\left({\frac {1}{\theta }}\right)}}}\left(\gamma ^{2}-1\right)^{{\frac {n}{2}}-1}\gamma \,\mathrm {d} \gamma }

Note that
can be shown to coincide with the thermodynamic definition of temperature.
Also useful is the expression of the distribution in the velocity space.[7] Given that
, one has:
![{\displaystyle {\begin{aligned}\mathrm {d} p_{1}\cdots \mathrm {d} p_{n}=p^{n-1}\,\mathrm {d} p\,\mathrm {d} \Omega _{n}&=\left(mc\right)^{n}\gamma ^{n-1}\beta ^{n-1}{\frac {\mathrm {d} (\beta \gamma )}{\mathrm {d} \beta }}\,\mathrm {d} \beta \,\mathrm {d} \Omega _{n}\\[1ex]&=\left(mc\right)^{n}\gamma ^{n+2}\beta ^{n-1}\,\mathrm {d} \beta \,\mathrm {d} \Omega _{n}\\[1ex]&=\left(mc\right)^{n}\gamma ^{n+2}\,\mathrm {d} \beta _{1}\cdots \mathrm {d} \beta _{n}\end{aligned}}}](./_assets_/eb734a37dd21ce173a46342d1cc64c92/0ba4b1b26df512419a4e96f6ccf7f3b8d75bacca.svg)
Hence
;\theta )\,\mathrm {d} \beta _{1}\cdots \mathrm {d} \beta _{n}=\pi ^{\frac {1-n}{2}}2^{-{\frac {n+1}{2}}}\,\theta ^{\frac {1-n}{2}}{\frac {e^{-\gamma (\beta )/\theta }}{\operatorname {K} _{\frac {n+1}{2}}{\!\!\left({\frac {1}{\theta }}\right)}}}\,\gamma (\beta )^{n+2}\,\mathrm {d} \beta _{1}\cdots \mathrm {d} \beta _{n}}

Take
(the "classic case" in our world):
;\theta )\,\mathrm {d} \gamma &={\frac {e^{-\gamma /\theta }\left(\gamma ^{2}-1\right)^{1/2}}{\theta \,\operatorname {K} _{2}{\!\left(1/\theta \right)}}}\,\gamma \,\mathrm {d} \gamma ,\\[1ex]f_{\text{MJ}}(\beta ;\theta )\,\mathrm {d} \beta _{1}\,\mathrm {d} \beta _{2}\,\mathrm {d} \beta _{3}&={\frac {e^{-\gamma (\beta )/\theta }\gamma (\beta )^{5}}{\theta \operatorname {K} _{2}{\!\left(1/\theta \right)}}}\,{\frac {\mathrm {d} \beta _{1}\,\mathrm {d} \beta _{2}\,\mathrm {d} \beta _{3}}{4\pi }}.\end{aligned}}}
![{\displaystyle {\begin{aligned}f_{\text{MJ}}(p;\theta )\,\mathrm {d} p_{1}\,\mathrm {d} p_{2}\,\mathrm {d} p_{3}&={\frac {e^{-{\gamma (\mathbf {p} )/\theta }}}{\theta \,\operatorname {K} _{2}{\!\left(1/\theta \right)}}}\,{\frac {\mathrm {d} p_{1}\,\mathrm {d} p_{2}\,\mathrm {d} p_{3}}{4\pi \left(mc\right)^{3}}},\\[1ex]f_{\text{MJ}}(\gamma ;\theta )\,\mathrm {d} \gamma &={\frac {e^{-\gamma /\theta }\left(\gamma ^{2}-1\right)^{1/2}}{\theta \,\operatorname {K} _{2}{\!\left(1/\theta \right)}}}\,\gamma \,\mathrm {d} \gamma ,\\[1ex]f_{\text{MJ}}(\beta ;\theta )\,\mathrm {d} \beta _{1}\,\mathrm {d} \beta _{2}\,\mathrm {d} \beta _{3}&={\frac {e^{-\gamma (\beta )/\theta }\gamma (\beta )^{5}}{\theta \operatorname {K} _{2}{\!\left(1/\theta \right)}}}\,{\frac {\mathrm {d} \beta _{1}\,\mathrm {d} \beta _{2}\,\mathrm {d} \beta _{3}}{4\pi }}.\end{aligned}}}](./_assets_/eb734a37dd21ce173a46342d1cc64c92/6fcd4fea6ca243390591e804e26e2f6934ec77b3.svg)
Note that when the MB distribution clearly deviates from the MJ distribution of the same temperature and dimensionality, one can misinterpret and deduce a different MB distribution that will give a good approximation to the MJ distribution. This new MB distribution can be either:
- a convected MB distribution, that is, an MB distribution with the same dimensionality, but with different temperature
and bulk speed
(or bulk energy
)
- an MB distribution with the same bulk speed, but with different temperature
and degrees of freedom
. These two types of approximations are illustrated.
Other properties
The MJ probability density function is given by:

This means that a relativistic non-quantum particle with parameter
has a probability of
of having its Lorentz factor in the interval
.
Cumulative distribution function
The MJ cumulative distribution function is given by:

That has a series expansion at
:

By definition
, regardless of the parameter
.
Average speed
To find the average speed,
, one must compute
, where
is the speed in terms of its Lorentz factor.
The integral simplifies to the closed-form expression:

This closed formula for
has a series expansion at
:

Or substituting the definition for the parameter
:

Where the first term of the expansion, which is independently of
, corresponds to the average speed in the Maxwell–Boltzmann distribution,
, whilst the following are relativistic corrections.
This closed formula for
has a series expansion at
:

Or substituting the definition for the parameter
:

Where it follows that
is an upper limit to the particle's speed, something only present in a relativistic context, and not in the Maxwell–Boltzmann distribution.