The Maxwell-Boltzmann Speed Distribution

Open a valve on a flask of nitrogen gas. Every molecule inside seems identical. It is not. Some molecules crawl at 100 m/s. Others streak past at 1500 m/s. This spread is not noise — it is a precise, temperature-dependent law. The Maxwell-Boltzmann speed distribution tells you exactly what fraction of molecules move within any speed interval. This holds at any temperature, for any molar mass. This distribution is the bridge between the microscopic chaos of molecular motion and macroscopic, measurable quantities. These include pressure, effusion rates, and reaction rates. It explains why light gases like helium escape a balloon faster than heavy ones. It explains why reaction rates rise steeply with temperature, even though average kinetic energy rises only linearly. It also explains why a small fraction of very fast molecules dominates chemistry at the high-energy tail. This article derives the distribution from the Boltzmann distribution over velocity components. It builds this up into a speed distribution. It then works through two fully numeric examples. You will see the derivation. Then you will see how the curve depends on mass and temperature. Finally, you will connect it to effusion and reaction kinetics.

Key takeaways

The Maxwell-Boltzmann distribution gives the probability density $f(v)$ for molecular speed $v$ in a gas at temperature $T$ .
It follows from assuming each velocity component ( $v_x$ , $v_y$ , $v_z$ ) is independently Gaussian-distributed (Boltzmann-weighted kinetic energy).
The distribution is not symmetric: it has a hard floor at $v = 0$ and a long tail toward high speeds.
Heavier molecules and lower temperatures shift the peak to lower speeds and narrow the curve; light molecules and high temperatures broaden it and shift it right.
Three characteristic speeds emerge from $f(v)$ : most probable $v_p$ , mean $\langle v \rangle$ , and root-mean-square $v_{rms}$ , with $v_p < \langle v \rangle < v_{rms}$ .
Key idea: molecular speeds are not fixed but statistically distributed according to a competition between the growing phase space at high speed ( $v^2$ ) and the shrinking Boltzmann weight ( $e^{-mv^2/2k_BT}$ ).

The core insight

The Maxwell-Boltzmann speed distribution arises because the number of ways a molecule can have speed $v$ (proportional to the surface area of a sphere of radius $v$ in velocity space, $\propto v^2$ ) is multiplied by the exponentially decaying Boltzmann probability of that kinetic energy, $e^{-mv^2/2k_BT}$ ; the product of a rising $v^2$ term and a falling exponential produces a distribution with a single peak, a zero at $v=0$ , and a tail extending to arbitrarily high speed.

The physical picture

Consider a gas of $N$ molecules, each with mass $m$ , in thermal equilibrium at temperature $T$ . Kinetic Molecular Theory (KMT) already tells you that pressure arises from molecular collisions with the walls, and that average translational kinetic energy is fixed by temperature alone: $\langle \tfrac{1}{2}mv^2 \rangle = \tfrac{3}{2}k_BT$ . But KMT as usually first presented only gives you the average — it says nothing about how individual molecular speeds are scattered around that average. The resolution starts one level down, at the level of velocity components. Each molecule has a velocity vector $\vec{v} = (v_x, v_y, v_z)$ . Ludwig Boltzmann's fundamental postulate is that in thermal equilibrium, the probability of any microstate is proportional to $e^{-E/k_BT}$ . Since the three Cartesian components of translational kinetic energy are independent and additive,

E = \tfrac{1}{2}mv_x^2 + \tfrac{1}{2}mv_y^2 + \tfrac{1}{2}mv_z^2,

the joint probability density factors into three identical, independent Gaussian distributions — one per Cartesian direction. This is the deep reason the speed distribution takes the shape it does: it is a three-dimensional Gaussian in velocity space, re-expressed in spherical (speed and direction) coordinates. When you convert from Cartesian velocity space $(v_x, v_y, v_z)$ to a speed-only description, you must count all the directions a molecule can point while having a given speed $v$ . That count is the surface area of a sphere of radius $v$ , which grows as $v^2$ . Multiplying this growing "phase space volume" by the shrinking Gaussian weight $e^{-mv^2/2k_BT}$ produces a curve that starts at zero, rises to a single maximum, and decays smoothly at high speed — the shape shown below.

Figure

The Maxwell–Boltzmann speed distribution at 100 K, 300 K, and 1000 K for N₂ gas M = 28.0 g/mol

The core relations

f(v) = 4\pi n \left( \frac{m}{2\pi k_B T} \right)^{3/2} v^2 e^{-mv^2/2k_BT}

where: $f(v)\,dv$ is the fraction of molecules with speed between $v$ and $v+dv$ , $m$ is the mass of one molecule, $k_B$ is Boltzmann's constant, and $n$ is a normalization constant (often written 1, since $f(v)$ is already a normalized probability density).

v_p = \sqrt{\frac{2RT}{M}}

where: $v_p$ is the most probable speed (the peak of $f(v)$ ), $R$ is the gas constant, and $M$ is the molar mass.

\langle v \rangle = \sqrt{\frac{8RT}{\pi M}}

where: $\langle v \rangle$ is the mean speed, averaged over the full distribution.

v_{rms} = \sqrt{\frac{3RT}{M}}

where: $v_{rms} = \sqrt{\langle v^2 \rangle}$ is the root-mean-square speed, directly tied to $\langle E_k \rangle = \tfrac{3}{2}k_BT$ .

v_p : \langle v \rangle : v_{rms} = \sqrt{2} : \sqrt{8/\pi} : \sqrt{3} \approx 1.000 : 1.128 : 1.225

where: this fixed ratio holds for every gas at every temperature — only the absolute scale changes with $M$ and $T$ .

Derivation AFrom Boltzmann-weighted velocity components to the speed distribution

Goal: derive $f(v)$ starting from the Boltzmann distribution over the three velocity components, and extract $v_p$ .

Step 1: Write the Boltzmann distribution for one velocity component. Because kinetic energy separates additively into $x$ , $y$ , $z$ contributions, each Cartesian velocity component is normally distributed:

g(v_x)\,dv_x = \left( \frac{m}{2\pi k_BT} \right)^{1/2} e^{-mv_x^2/2k_BT}\,dv_x,

and identically for $g(v_y)$ , $g(v_z)$ . This is simply a Gaussian with variance $k_BT/m$ , a direct consequence of the equipartition result $\langle \tfrac{1}{2}mv_x^2\rangle = \tfrac{1}{2}k_BT$ from the molecular interpretation of temperature.

Step 2: Build the joint velocity-space probability density. Because $v_x$ , $v_y$ , $v_z$ are statistically independent, the joint density is the product:

G(\vec v) = g(v_x)g(v_y)g(v_z) = \left( \frac{m}{2\pi k_BT} \right)^{3/2} e^{-m(v_x^2+v_y^2+v_z^2)/2k_BT}.

Note that $v_x^2+v_y^2+v_z^2 = v^2$ , the squared speed — so $G(\vec v)$ depends only on the magnitude of $\vec v$ , not its direction. The gas is isotropic, as physically required.

Step 3: Convert to spherical coordinates and integrate out direction. The probability of finding a molecule with velocity vector in a small volume element $d^3v = v^2 \sin\theta \, dv \, d\theta \, d\phi$ is $G(\vec v)\, d^3v$ . To get the probability of speed lying between $v$ and $v+dv$ regardless of direction, integrate over all solid angle:

f(v)\,dv = G(\vec v)\, v^2\, dv \int_0^{2\pi}\!\!\int_0^{\pi} \sin\theta \, d\theta \, d\phi = 4\pi v^2\, G(\vec v)\, dv.

The factor $4\pi v^2$ is exactly the surface area of a sphere of radius $v$ in velocity space — this is where the "phase-space volume grows as $v^2$ " idea becomes precise.

Step 4: Assemble the full speed distribution. Substituting $G(\vec v)$ from Step 2:

f(v) = 4\pi \left( \frac{m}{2\pi k_BT} \right)^{3/2} v^2 e^{-mv^2/2k_BT}.

This is normalized: $\int_0^\infty f(v)\,dv = 1$ , since the Gaussian normalization in Step 1 was already exact in each of the three directions.

Step 5: Locate the peak (most probable speed). Differentiate $f(v)$ with respect to $v$ and set the result to zero. Writing $a = m/2k_BT$ for brevity,

\frac{d}{dv}\left( v^2 e^{-av^2} \right) = 2v\,e^{-av^2} - 2av^3 e^{-av^2} = 2v\,e^{-av^2}\left(1 - av^2\right).

Setting the bracket to zero (the $v=0$ root is a minimum, not the peak) gives $av_p^2 = 1$ , so

v_p^2 = \frac{1}{a} = \frac{2k_BT}{m}.

Multiplying numerator and denominator by Avogadro's number $N_A$ converts to molar quantities ( $N_Ak_B=R$ , $N_Am=M$ ):

\boxed{v_p = \sqrt{\frac{2RT}{M}}}

The mean speed $\langle v \rangle = \int_0^\infty v f(v)\, dv$ and $\langle v^2\rangle = \int_0^\infty v^2 f(v)\,dv$ follow from standard Gaussian moment integrals, yielding the $\sqrt{8RT/\pi M}$ and $3RT/M$ results quoted above.

Worked examples

Example 1 — Foundational: characteristic speeds of N₂ at room temperature. Find $v_p$ , $\langle v \rangle$ , and $v_{rms}$ for nitrogen gas, $M = 28.02\times10^{-3}$ kg/mol, at $T = 298$ K. Use $R = 8.314$ J/(mol·K). Most probable speed:

v_p = \sqrt{\frac{2RT}{M}} = \sqrt{\frac{2(8.314)(298)}{28.02\times10^{-3}}} = \sqrt{\frac{4955.1}{0.02802}} = \sqrt{1.7684\times10^5} = 420.5 \text{ m/s}.

Mean speed:

\langle v \rangle = \sqrt{\frac{8RT}{\pi M}} = \sqrt{\frac{8(8.314)(298)}{\pi (0.02802)}} = \sqrt{\frac{19820.5}{0.08803}} = \sqrt{2.2515\times10^5} = 474.5 \text{ m/s}.

Root-mean-square speed:

v_{rms} = \sqrt{\frac{3RT}{M}} = \sqrt{\frac{3(8.314)(298)}{0.02802}} = \sqrt{\frac{7432.7}{0.02802}} = \sqrt{2.6526\times10^5} = 515.0 \text{ m/s}.

Sanity check: the ratio $v_p : \langle v \rangle : v_{rms}$ should be $1 : 1.128 : 1.225$ . Here $474.5/420.5 = 1.128$ ✓ and $515.0/420.5 = 1.225$ ✓. These numbers also match the familiar textbook value that N₂ molecules at room temperature move at roughly "the speed of a jet plane," several hundred m/s — much faster than everyday intuition suggests, because momentum, not speed, is what we perceive as "fast" for large objects.

Example 2 — Applied: comparing helium and effusion, plus fraction of "fast" molecules. (a) Compare $v_{rms}$ for He ( $M = 4.003\times10^{-3}$ kg/mol) and N₂ at 298 K, and relate the ratio to Graham's law of effusion.

v_{rms}(\text{He}) = \sqrt{\frac{3(8.314)(298)}{4.003\times10^{-3}}} = \sqrt{\frac{7432.7}{0.004003}} = \sqrt{1.8568\times10^{6}} = 1362.6 \text{ m/s}.

Ratio to N₂:

\frac{v_{rms}(\text{He})}{v_{rms}(\text{N}_2)} = \frac{1362.6}{515.0} = 2.646.

Graham's law predicts the effusion rate ratio equals $\sqrt{M(\text{N}_2)/M(\text{He})} = \sqrt{28.02/4.003} = \sqrt{7.000} = 2.646$ — an exact match, since both results trace back to the same $1/\sqrt{M}$ scaling embedded in $f(v)$ . This is why helium leaks out of a balloon roughly 2.6 times faster than nitrogen would, at the same temperature. (b) Estimate, order-of-magnitude, why raising temperature from 298 K to 500 K dramatically increases the fraction of "reactive" N₂ molecules exceeding a threshold speed of 1000 m/s, even though $v_{rms}$ itself only rises modestly. At 298 K: $v_{rms} = 515.0$ m/s (from above), so 1000 m/s is roughly $1000/515.0 = 1.94 \times v_{rms}$ — deep in the tail of $f(v)$ . At 500 K:

v_{rms}(500\text{ K}) = \sqrt{\frac{3(8.314)(500)}{0.02802}} = \sqrt{\frac{12471}{0.02802}} = \sqrt{4.4507\times10^5} = 667.1 \text{ m/s},

so 1000 m/s is now only $1000/667.1 = 1.50 \times v_{rms}$ — much closer to the bulk of the distribution. Because the high-speed tail of $f(v)$ falls off as $e^{-mv^2/2k_BT}$ , moving the threshold from $1.94\,v_{rms}$ to $1.50\,v_{rms}$ (i.e., closer to the peak, in relative terms) increases the tail population by roughly an order of magnitude even though $v_{rms}$ itself rose only 30% (515 → 667 m/s). This exponential sensitivity of the high-speed tail to $T$ is the molecular-level origin of the Arrhenius law: reaction rates rise steeply with temperature because they sample this shrinking-then-growing tail of $f(v)$ , not because average speeds change much.

Limiting cases & common traps

Situation	What happens to $f(v)$	Why
$T \to 0$	Distribution collapses to a spike at $v=0$	Thermal energy $k_BT \to 0$ removes all spread
$T$ increases	Peak shifts right ( $v_p \propto \sqrt{T}$ ), curve broadens and flattens	More thermal energy available per molecule
$M$ increases (fixed $T$ )	Peak shifts left, curve narrows	Heavier molecules need less speed for the same $k_BT$
$M$ decreases (fixed $T$ )	Peak shifts right, curve broadens	Light molecules move faster on average
Area under any segment of $f(v)$	Always equals the fraction of molecules in that speed range	$f(v)$ is a probability density, not a probability

The chemist behind it

James Clerk Maxwell first derived the distribution in 1860 using a strikingly modern argument: he assumed the three velocity components must be statistically independent and that the distribution could only depend on speed (isotropy), and showed this forces a Gaussian form for each component. Ludwig Boltzmann later generalized the argument using the full statistical-mechanical framework of the Boltzmann distribution, connecting it rigorously to entropy and the approach to equilibrium via his H-theorem. The distribution is jointly named for both, reflecting Maxwell's original insight and Boltzmann's deeper statistical foundation.

Origins · 1 minLudwig Boltzmann and how this was discovered

Ludwig Boltzmann

1844–1906 · Austrian physicist who gave the Maxwell speed distribution its rigorous statistical-mechanical foundation and connected molecular motion to entropy via S = k_B ln W

Ask the tutor: How does this speed distribution translate into the collision frequency and mean free path a molecule experiences in a real gas?

Why is the speed distribution not just a simple bell curve?

A bell curve (Gaussian) is exactly what you get for a single velocity component, like $v_x$ , which can be positive or negative. Speed, however, is the magnitude of the velocity vector and can never be negative. Converting from Cartesian components to speed introduces the $4\pi v^2$ geometric factor from spherical coordinates, which forces $f(v)=0$ at $v=0$ and skews the whole curve rightward, producing the asymmetric shape you see in the plot rather than a symmetric bell.

Does every gas have a different-shaped distribution?

The functional form of $f(v)$ is identical for every ideal gas; only the two parameters $m$ and $T$ change. If you plot $f(v)$ against the reduced variable $v/v_p$ , every gas at every temperature collapses onto exactly the same universal curve. This is why the fixed ratio $v_p:\langle v\rangle:v_{rms} \approx 1:1.128:1.225$ holds universally — it is a property of the mathematical shape, not of any specific molecule.

How is this distribution actually measured experimentally?

Classic experiments (Zartman 1931, Ko 1934, and refined versions since) use rotating slotted-disk velocity selectors: a beam of molecules effusing from an oven passes through a rotating drum with a helical slot, and only molecules with a speed matching the drum's rotation rate pass through to a detector. By varying rotation speed and measuring detector flux, experimenters directly reconstruct $f(v)$ , and the measured curves match Maxwell's 1860 prediction to high precision.

How does this distribution connect to reaction rates?

Only molecules with kinetic energy exceeding the activation energy $E_a$ can react upon collision, and that population is exactly the high-speed tail of $f(v)$ . Because the tail scales as $e^{-mv^2/2k_BT}$ , which is equivalent to $e^{-E_a/RT}$ when converted to energy, the fraction of sufficiently energetic molecules — and hence the reaction rate — depends exponentially on temperature. This is precisely the physical content of the Arrhenius equation.

What is the difference between the Maxwell-Boltzmann speed and energy distributions?

The speed distribution $f(v)$ and the closely related kinetic energy distribution are two views of the same physics, related by $E = \tfrac{1}{2}mv^2$ . The energy distribution is often more directly useful in reaction kinetics because activation energies are quoted in energy units, while the speed distribution is more natural for transport properties like effusion, diffusion, and collision frequency, covered next.