Maxwell-Boltzmann Speed Distribution

Not every molecule in a gas moves at the same speed. At any instant some crawl, most bustle along near a typical pace, and a rare few streak far faster than average. The Maxwell–Boltzmann speed distribution is the exact statistical law governing this spread: the probability that a molecule of mass $m$ in a gas at temperature $T$ has speed near $v$ is $f(v)=4\pi\big(m/2\pi k_BT\big)^{3/2}v^2 e^{-mv^2/2k_BT}$ . It is a single, universal curve — bell-shaped but skewed, rising from zero, peaking at the most probable speed, and trailing off in an exponential tail. The shape is the product of two competing factors. The Boltzmann factor $e^{-mv^2/2k_BT}$ penalizes high energies, pushing molecules toward low speed. Against it, the geometric factor $v^2$ counts how many directions correspond to a given speed — the surface area of a sphere of radius $v$ in velocity space. Slow molecules have almost no directions available (a tiny sphere), so $f(0)=0$ ; fast molecules face the exponential guillotine. The winner is a peak at intermediate speed, the most probable speed $v_p=\sqrt{2k_BT/m}$ . Maxwell derived this distribution in 1860 from a symmetry argument alone — that the three velocity components are independent and isotropic — years before Boltzmann grounded it in the general canonical ensemble. It was one of the first triumphs of the statistical view of matter, and it remains indispensable: it fixes reaction rates (only the fast tail reacts), effusion and evaporation rates, Doppler line broadening in spectroscopy, and the escape of light gases from planetary atmospheres. Everything a gas does that depends on molecular speed traces back to this curve.

Key takeaways

The speed distribution is $f(v)=4\pi\big(m/2\pi k_BT\big)^{3/2}v^2 e^{-mv^2/2k_BT}$ , normalized so $\int_0^\infty f(v)\,dv=1$ .
Its shape is the product of the Boltzmann factor $e^{-mv^2/2k_BT}$ (kills high speeds) and the velocity-space shell area $4\pi v^2$ (kills zero speed), so $f(0)=0$ and there is an interior peak.
The most probable speed is $v_p=\sqrt{2k_BT/m}$ , where $df/dv=0$ .
The distribution has moments giving the mean speed $\bar v=\sqrt{8k_BT/\pi m}$ and the rms speed $v_{\rm rms}=\sqrt{3k_BT/m}$ , always ordered $v_p<\bar v<v_{\rm rms}$ .
It scales as $\sqrt{T/m}$ : hotter or lighter molecules are faster, and the whole curve stretches; the high-speed tail controls reaction rates and atmospheric escape.
Key idea: the speed distribution is a tug-of-war between the exponential Boltzmann penalty on energy and the $v^2$ growth in the number of available directions — their product is a skewed bell peaked at $v_p=\sqrt{2k_BT/m}$ .

The core insight

The Maxwell–Boltzmann speed distribution $f(v)=4\pi(m/2\pi k_BT)^{3/2}v^2 e^{-mv^2/2k_BT}$ is the product of two factors: the geometric $v^2$ (the surface area of a sphere of radius $v$ in velocity space, so many directions share a large speed) and the Boltzmann exponential (energetic penalty). The first forces $f(0)=0$ ; the second cuts off the tail; together they peak at the most probable speed $v_p=\sqrt{2k_BT/m}$ .

Where this lives today

The Maxwell speed distribution is a working tool wherever molecular speeds matter. In chemical kinetics it underlies the Arrhenius law: only molecules in the high-speed tail carry enough energy to react, so reaction rates are exquisitely sensitive to temperature through the tail. In spectroscopy it sets the Doppler width of atomic and molecular lines — the spread of speeds along the line of sight broadens every transition, and measuring that width is a standard gas thermometer. Vacuum technology uses it for effusion and gas-flow rates through apertures. Atmospheric science uses the tail to compute Jeans escape of light gases from planets and moons. Even in plasma physics and astrophysics the Maxwellian is the default equilibrium distribution against which non-thermal tails (from shocks, turbulence, or acceleration) are measured. And in cold-atom laboratories, imaging the expansion of a released cloud reconstructs its velocity distribution directly, using the Maxwellian as the classical benchmark that Bose or Fermi statistics deviate from.

The physical picture

Imagine every molecule's velocity as an arrow from the origin in a three-dimensional velocity space. The Maxwell–Boltzmann velocity distribution says the tips of these arrows form an isotropic Gaussian cloud centred on the origin, densest at zero velocity and thinning outward. But when you ask about speed — the length of the arrow, ignoring direction — you must gather all arrows of a given length, and those form a spherical shell of radius $v$ . The number of arrows in that shell is the local density (the Gaussian, $\propto e^{-mv^2/2k_BT}$ ) times the shell's surface area ( $4\pi v^2$ ). This is why the speed distribution vanishes at $v=0$ even though the velocity distribution is largest there: a sphere of zero radius has zero area, so no matter how dense the cloud is at the origin, there are no directions to a zero-length arrow. As $v$ grows the shell area swells, but the Gaussian density collapses; the product peaks at an intermediate speed.

Figure

The speed distribution f(v) is the product of the falling Boltzmann factor e^{-mv²/2kT} and the rising shell area 4πv²: their product is a skewed bell peaked at the most probable speed v_p, with a long high-speed tail.

The punchline figure places the three characteristic speeds directly on the curve. They are not equal — the skew of the distribution guarantees $v_p<\bar v<v_{\rm rms}$ — and the gap between them is the visible fingerprint of the long high-speed tail. It is that tail, not the peak, that decides which molecules react, evaporate, or escape a planet's gravity.

Figure

Punchline: the three characteristic speeds marked ON the f(v) curve — most probable v_p (peak), mean v̄, and root-mean-square v_rms — always ordered v_p < v̄ < v_rms because the distribution is skewed toward high speed.

The core equations

The speed distribution is built by taking the isotropic Gaussian velocity distribution and integrating over all directions at fixed speed, which introduces the spherical shell factor $4\pi v^2$ .

f(v)=4\pi\left(\frac{m}{2\pi k_BT}\right)^{3/2}v^2\,e^{-mv^2/2k_BT},\qquad \int_0^\infty f(v)\,dv=1 \qquad (1)

where: $m$ — molecular mass; $T$ — temperature; the prefactor is fixed by normalization; the $v^2$ is the velocity-space shell area (divided by the $4\pi$ already shown), and the exponential is the Boltzmann weight of the kinetic energy $\tfrac12mv^2$ . The most probable speed is where the curve peaks — where the rising $v^2$ and falling exponential balance, $df/dv=0$ .

v_p=\sqrt{\frac{2k_BT}{m}} \qquad (2)

where: $v_p$ — most probable speed, the location of the maximum. Setting $\tfrac{d}{dv}\big(v^2e^{-mv^2/2k_BT}\big)=0$ gives $2v=mv^3/k_BT$ , hence $v_p^2=2k_BT/m$ . It is the natural velocity scale of the distribution. The moments of $f(v)$ give the other two benchmark speeds. Each is a Gaussian-type integral evaluated with the gamma function.

\bar v=\int_0^\infty v\,f(v)\,dv=\sqrt{\frac{8k_BT}{\pi m}},\qquad v_{\rm rms}=\sqrt{\langle v^2\rangle}=\sqrt{\frac{3k_BT}{m}} \qquad (3)

where: $\bar v$ — mean speed (first moment); $v_{\rm rms}$ — root-mean-square speed (square root of the second moment). Because $\langle v^2\rangle>\langle v\rangle^2$ , we have $\bar v<v_{\rm rms}$ , and both exceed the peak $v_p$ . Their ratios are pure numbers, independent of $m$ and $T$ — a fingerprint of the distribution's shape.

v_p:\bar v:v_{\rm rms}=\sqrt2:\sqrt{8/\pi}:\sqrt3=1.414:1.596:1.732 \qquad (4)

where: the ratios are set entirely by the functional form $v^2e^{-mv^2/2k_BT}$ ; they are the same for every gas at every temperature. This universality is why the ordering $v_p<\bar v<v_{\rm rms}$ never changes.

Derivation ANormalization of the speed distribution via the gamma function

The goal is to prove that $f(v)$ as written in (1) is correctly normalized, $\int_0^\infty f(v)\,dv=1$ , and in doing so to expose where the prefactor comes from and why the shell factor is $4\pi v^2$ . The strategy is to reduce the speed integral to a standard gamma-function integral $\int_0^\infty v^2 e^{-av^2}\,dv$ and check that the prefactor is exactly the reciprocal.

Step 1. Start from the velocity distribution, an isotropic Gaussian $f_{\rm vel}(\mathbf v)=(m/2\pi k_BT)^{3/2}e^{-mv^2/2k_BT}$ , normalized over all of velocity space $\int d^3v\,f_{\rm vel}=1$ .

Step 2. Convert to speed by integrating over directions at fixed $|\mathbf v|=v$ : the volume element $d^3v=v^2\,dv\,d\Omega$ , and integrating $d\Omega$ over the full sphere gives $4\pi$ . Hence $f(v)=f_{\rm vel}(v)\cdot4\pi v^2$ , which is exactly (1). Why the $v^2$ . It is the surface area of a sphere of radius $v$ in velocity space (up to the $4\pi$ ): many velocity directions share the same speed, and there are more of them the larger $v$ is. This geometric factor is the entire reason the speed distribution differs in shape from the velocity distribution.

Step 3. Now do the normalization integral. Write $a=m/2k_BT$ so $f(v)=4\pi(a/\pi)^{3/2}v^2 e^{-av^2}$ . The needed integral is $\int_0^\infty v^2 e^{-av^2}\,dv=\tfrac14\sqrt{\pi/a^3}=\tfrac14\sqrt\pi\,a^{-3/2}$ .

Step 4. Assemble: $\int_0^\infty f\,dv=4\pi(a/\pi)^{3/2}\cdot\tfrac14\sqrt\pi\,a^{-3/2}=\pi\cdot\pi^{-3/2}\cdot\sqrt\pi=1$ . The prefactor and the integral cancel exactly. Symbolic integration of (1) gives $\int_0^\infty f(v)\,dv=1$ exactly, and a numeric trapezoidal check for N₂ at 300 K returns 1.000000.

Go deeper · Derivation B · 1 minThe most probable, mean, and rms speeds and their ordering

Derivation BThe most probable, mean, and rms speeds and their ordering

The goal is to compute all three characteristic speeds from $f(v)$ , establish the exact ratios of (4), and prove the ordering $v_p<\bar v<v_{\rm rms}$ holds universally. The strategy is one maximization (for $v_p$ ) and two gamma-function moment integrals (for $\bar v$ and $v_{\rm rms}$ ), then compare the pure-number coefficients.

Step 1. Most probable speed: maximize $v^2 e^{-mv^2/2k_BT}$ . Differentiate and set to zero: $2v-mv^3/k_BT=0$ , so $v_p^2=2k_BT/m$ , i.e. $v_p=\sqrt{2k_BT/m}$ .

Step 2. Mean speed: $\bar v=\int_0^\infty v\,f\,dv$ . The integral $\int_0^\infty v^3 e^{-av^2}\,dv=1/(2a^2)$ leads, after inserting the prefactor, to $\bar v=\sqrt{8k_BT/\pi m}$ .

Step 3. Mean-square speed: $\langle v^2\rangle=\int_0^\infty v^2 f\,dv$ . The integral $\int_0^\infty v^4 e^{-av^2}\,dv=\tfrac38\sqrt\pi\,a^{-5/2}$ gives $\langle v^2\rangle=3k_BT/m$ , so $v_{\rm rms}=\sqrt{3k_BT/m}$ . (This also confirms equipartition: $\tfrac12m\langle v^2\rangle=\tfrac32k_BT$ .)

Step 4. Form the ratios in units of $\sqrt{k_BT/m}$ : $v_p=\sqrt2\approx1.414$ , $\bar v=\sqrt{8/\pi}\approx1.596$ , $v_{\rm rms}=\sqrt3\approx1.732$ . Why the ordering is fixed. The coefficients $\sqrt2<\sqrt{8/\pi}<\sqrt3$ are pure numbers with no $m$ or $T$ , so $v_p<\bar v<v_{\rm rms}$ for every gas at every temperature — the skew of the distribution is a permanent feature.

Step 5. The ratios $\bar v/v_p=2/\sqrt\pi\approx1.128$ and $v_{\rm rms}/v_p=\sqrt{3/2}\approx1.225$ are the compact, temperature-independent signatures often quoted. Symbolic moments confirm $v_p=\sqrt{2k_BT/m}$ , $\bar v=\sqrt{8k_BT/\pi m}$ , $v_{\rm rms}=\sqrt{3k_BT/m}$ with $\langle v^2\rangle=3k_BT/m$ , numeric coefficients $1.4142<1.5958<1.7321$ , and $df/dv|_{v_p}=0$ .

Limiting cases

The distribution scales rigidly with $\sqrt{T/m}$ ; changing $T$ or $m$ stretches or compresses it without altering its shape.

Case	Effect	Meaning
Higher temperature $T$	Curve broadens and shifts right ( $\propto\sqrt T$ )	Faster, more spread-out speeds; longer tail.
Lighter molecules (small $m$ )	Curve shifts right ( $\propto1/\sqrt m$ )	Light gases (H₂, He) far faster than heavy ones at the same $T$ .
$v\to0$	$f(v)\to0$ (as $v^2$ )	No directions to a zero-length velocity; slow molecules are rare.
$v\to\infty$	Exponential decay $e^{-mv^2/2k_BT}$	The tail that drives reactions, evaporation, atmospheric escape.
Relativistic / very high $T$	Maxwell–Jüttner distribution	The non-relativistic form breaks down when $k_BT\sim mc^2$ .

Worked examples

Example 1 — The three speeds of nitrogen at room temperature (enumerate, apply, boundary number). Compute $v_p$ , $\bar v$ , and $v_{\rm rms}$ for N₂ ( $m=28.0\,\mathrm{u}=4.65\times10^{-26}\,\mathrm{kg}$ ) at $T=300\,\mathrm{K}$ .

Step 1. Enumerate the scale $\sqrt{k_BT/m}=\sqrt{(1.381\times10^{-23})(300)/(4.65\times10^{-26})}=\sqrt{8.91\times10^{4}}=298.5\,\mathrm{m/s}$ .

Step 2. Apply the coefficients: $v_p=\sqrt2\times298.5=\mathbf{422\,m/s}$ ; $\bar v=\sqrt{8/\pi}\times298.5=1.596\times298.5=\mathbf{476\,m/s}$ ; $v_{\rm rms}=\sqrt3\times298.5=1.732\times298.5=\mathbf{517\,m/s}$ .

Step 3. Check the ordering: $422<476<517$ — the universal $v_p<\bar v<v_{\rm rms}$ , as required.

Step 4. Boundary number: these are roughly the speed of sound in air ( $\approx343\,\mathrm{m/s}$ ) and comparable — molecular speeds set the speed of sound, since sound is a collective disturbance carried by molecular motion. The mean speed $476\,\mathrm{m/s}$ is about $1.4\times$ the sound speed, the difference being that sound is a directed, collective wave while $\bar v$ is a randomized single-molecule average.

Example 2 — Why helium escapes but nitrogen stays. Compare the rms speeds of helium and nitrogen at $1000\,\mathrm{K}$ (a temperature reached in the exosphere) and relate to escape.

Step 1. $v_{\rm rms}\propto1/\sqrt m$ , so $v_{\rm rms}(\mathrm{He})/v_{\rm rms}(\mathrm{N_2})=\sqrt{m_{\rm N_2}/m_{\rm He}}=\sqrt{28.0/4.00}=\sqrt7=\mathbf{2.65}$ .

Step 2. Numerically, $v_{\rm rms}(\mathrm{He},1000\,\mathrm{K})=\sqrt{3(1.381\times10^{-23})(1000)/(6.64\times10^{-27})}=\sqrt{6.24\times10^{6}}\approx\mathbf{2500\,m/s}$ , while N₂ manages only $\approx940\,\mathrm{m/s}$ .

Step 3. The escape velocity from Earth is $\approx11{,}200\,\mathrm{m/s}$ . No single molecule at these mean speeds escapes, but the tail of the distribution does: because helium's whole curve is $2.65\times$ faster, its exponential tail above the escape speed is vastly more populated. Over geological time this is why Earth has retained nitrogen and oxygen but lost nearly all its primordial helium and hydrogen — the high-speed tail of the Maxwell distribution, integrated over eons, has emptied the light gases into space.

Origins · 1 minmaxwell speed distribution and how this was discovered

maxwell speed distribution

kinetic-theory result, Maxwell 1860 · the speed law of a gas

The Maxwell speed distribution is the probability law for molecular speeds in a gas at thermal equilibrium, derived by James Clerk Maxwell in 1860 in his "Illustrations of the Dynamical Theory of Gases." Maxwell's argument was strikingly modern: he assumed only that the three Cartesian velocity components are statistically independent and that the distribution is isotropic (no preferred direction), and showed these force a Gaussian in each component — hence the $v^2e^{-mv^2/2k_BT}$ speed law. It was one of the first uses of a probability distribution in physics. Ludwig Boltzmann later generalized it to the distribution in an arbitrary potential, and the combined Maxwell–Boltzmann distribution became the cornerstone of kinetic theory, later confirmed directly by molecular-beam experiments (Stern, 1920s; Miller and Kusch, 1955).

History

1859

— Maxwell presents his first derivation of the velocity distribution to the British Association.

1860

— Maxwell's "Illustrations of the Dynamical Theory of Gases" publishes the speed distribution from independence and isotropy.

1868

— Boltzmann generalizes the distribution to particles in an external potential (the Maxwell–Boltzmann distribution).

1920

— Otto Stern's molecular-beam experiments provide the first direct measurement of molecular speeds.

1955

— Miller and Kusch precisely confirm the Maxwell speed distribution with a rotating-drum velocity selector.

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., §3.5 and Ch. 6 (kinetic theory) · Reif, Fundamentals of Statistical and Thermal Physics, §7.9–7.10 · Kardar, Statistical Physics of Particles, §3 · Maxwell, "Illustrations of the Dynamical Theory of Gases," Phil. Mag. (1860) · Landau & Lifshitz, Statistical Physics Part 1, §29. Key results: $\int_0^\infty f(v)\,dv=1$ symbolically and numerically ( $=1.000000$ for N₂ at 300 K); $df/dv=0$ at $v_p=\sqrt{2k_BT/m}$ ; $\bar v=\sqrt{8k_BT/\pi m}$ , $\langle v^2\rangle=3k_BT/m$ ; coefficients $\sqrt2:\sqrt{8/\pi}:\sqrt3=1.414:1.596:1.732$ ; numerics for N₂ at 300 K: $v_p=422$ , $\bar v=476$ , $v_{\rm rms}=517\,\mathrm{m/s}$ ; $f(v)=f_{\rm vel}\cdot4\pi v^2$ . Constants used: $k_B=1.380649\times10^{-23}\,\mathrm{J/K}$ , $1\,\mathrm{u}=1.66054\times10^{-27}\,\mathrm{kg}$ .

Builds on: Classical Ideal Gas · Maxwell-Boltzmann Velocity Distribution · Equipartition Theorem Leads to: Characteristic Speeds: Mean, RMS, and Most Probable · Thermal de Broglie Wavelength · Classical Limit of Quantum Statistics

Why is the speed distribution skewed rather than a symmetric bell?

Because it is the product of a symmetric-looking Gaussian in speed and the strictly rising $v^2$ shell factor. The $v^2$ pulls the distribution's weight toward higher speeds and forces $f(0)=0$ , while the exponential cuts off the far tail. The result leans right: there is a hard wall at $v=0$ but a long soft tail at high $v$ . This asymmetry is exactly why the mean speed $\bar v$ (sensitive to the tail) exceeds the most probable speed $v_p$ (the peak), and why $v_{\rm rms}$ (which weights the tail even more, by $v^2$ ) is larger still.

How does the distribution change with temperature?

The whole curve scales with $\sqrt T$ : raising $T$ moves every characteristic speed outward as $\sqrt T$ and broadens the distribution, lowering and flattening the peak (the area stays one). Physically the gas gets faster and its speeds more spread out. The effect on the high-speed tail is even more dramatic than on the peak, because the tail is exponential — a modest temperature rise can multiply the population above a threshold speed many-fold, which is precisely why reaction rates and evaporation rates climb so steeply with temperature.

Why is v_rms used for pressure and energy but v̄ for collisions?

Because different physical quantities weight the distribution differently. The pressure and kinetic energy involve $\langle v^2\rangle$ (energy $\propto v^2$ ), so $v_{\rm rms}$ is the relevant average — indeed $\tfrac12m\langle v^2\rangle=\tfrac32k_BT$ . Collision and effusion rates, by contrast, count how fast molecules travel and hit things on average, which is the first moment $\bar v$ . Using $v_{\rm rms}$ where $\bar v$ belongs (or vice versa) is a common source of $\sim10\%$ errors; always ask which power of $v$ the physics weights.

Does the distribution assume the molecules do not collide?

No — collisions are essential; they are what thermalizes the gas and drive it to the Maxwell–Boltzmann form in the first place. The distribution is the equilibrium reached after many collisions have randomized the speeds. What it does assume is that the gas is dilute enough that the molecules are independent between collisions (the ideal-gas assumption) and that equilibrium has been established. Boltzmann's H-theorem shows that collisions relax any initial speed distribution toward this Maxwellian, which is the unique collisional fixed point.

What replaces it at very high temperature or for photons?

When $k_BT$ becomes comparable to $mc^2$ the non-relativistic kinetic energy $\tfrac12mv^2$ is wrong and the distribution is replaced by the relativistic Maxwell–Jüttner distribution. For massless particles like photons there is no rest mass and the relevant equilibrium law is Planck's blackbody distribution, not Maxwell–Boltzmann. And when quantum degeneracy sets in ( $n\lambda_T^3\gtrsim1$ ), the classical Maxwell–Boltzmann distribution gives way to Bose–Einstein or Fermi–Dirac statistics. The Maxwellian is the non-relativistic, non-degenerate classical limit.

How was the distribution measured directly?

By molecular-beam experiments. A beam of atoms is let out of an oven through a slit, and a rotating slotted drum or a pair of synchronized rotating disks acts as a velocity selector: only molecules with a speed matching the rotation rate pass through to a detector. Sweeping the rotation rate maps out how many molecules have each speed — directly tracing $f(v)$ . Otto Stern pioneered this in the 1920s, and Miller and Kusch (1955) confirmed the Maxwell distribution to high precision, turning a theoretical probability law into a directly verified experimental fact.