Consider an ideal gas obeying the Maxwell-Boltzmann distribution i.e

$$f(v) = \bigg(\frac{m}{2 \pi k_{B} T}\bigg)^{3/2} \exp \left(-\frac{m v^{2}}{2 k_{B} T} \right) \, .$$

The probability distribution in 3D velocity space ($v^{2} = v_{x}^2 + v_{y}^2 + v_{z}^2$). How might you determine the average speed the particles are moving at, $\langle |v_{z}| \rangle$, in one direction?

Additionally if my ideal gas is now confined to a hemisphere in velocity space i.e we have the conditions $- \infty \leq v_{x}, v_{y} \leq \infty$ and $ 0 \leq v_{z} \leq \infty$ but it still has a Maxwell Boltzmann velocity distribution (except I think the normalization factor on $f(v)$ should change) then what is the average speed, or velocity, in the z direction $\langle v_{z} \rangle$, will this be the same as $|\langle v_{z}| \rangle$ from the previous answer?

  • $\begingroup$ What are your ideas as to what would happen? For example, what makes you think that $<|v_{z}|>$ will not change? $\endgroup$
    – Farcher
    Mar 22, 2017 at 8:12
  • $\begingroup$ Well I think it shouldn't change because of symmetry, if all the $v_{z} < 0$ have their signs flipped then that should have no effect on the average speed. Also I believe that $|<v_{z}>|$ can be determined by integrating out the x and y dependence of $f(v)$? $\endgroup$ Mar 22, 2017 at 9:11
  • $\begingroup$ It seems that your distribution should be that of $f(\vec{v})$, but not $f(v)$, which should carry a $v^2$ factor. $\endgroup$
    – velut luna
    Sep 10, 2017 at 16:13

2 Answers 2


The average speed of particles in a particular direction will always be smaller than the average speed of particles.

Imagine that you have three particles with components of their velocity $(1,1,1)\, \rm ms^{-1}$.
Their average speed is $<v>= \sqrt 3\, \rm ms^{-1}$ whilst their average speed in the x-direction is $<v_{\rm x}>=1\, \rm ms^{-1}$

So what you need to do is use the distribution of velocities in the x-direction

$$f(v_\rm{x}) = \left(\frac{m}{2 \pi k_{B} T}\right)^\frac{1}{2} \exp\left ({-\frac{m v_{\rm x}^{2}}{2 k_{B} T}}\right )$$

and do the following integration

$\displaystyle \int_0^\infty v_\rm{x}\, f(v_\rm{x}) \,dv_{\rm x}$ which will give you an answer of $\dfrac {<v>}{4}$ where $<v>$ is the avergae speed of the particles.

You may find these notes of use?

  • $\begingroup$ I think this would not be correct as this is not consistent with$<|v_{x}|>$ of the distribution, your integral $\int_{0}^{\infty}v_{x}f(v_{x})dv_{x}$ is not properly weighted by the sum of the probabilities in the $0 \leq v_{x} \leq \infty$ region. Hence it requires a factor of $2$ out the front. I asked this question because I have seen several people stating this result yet I am convinced $<v_{x}> = \frac{<v>}{2}$. where $<v_{x}>$ is the average speed in the x direction and $<v>$ is the average speed in all 3 directions $\endgroup$ May 1, 2017 at 15:33
  • $\begingroup$ V. Molecular flux lehman.edu/faculty/dgaranin/Statistical_Thermodynamics/… $\endgroup$
    – Farcher
    May 1, 2017 at 21:38
  • $\begingroup$ Yes that is for the molecular flux however which only accounts for the particles moving in one specific direction across the barrier, it does not tell you that $<v_{x}>= \frac{1}{4}<v>$. In fact, because only half the density of particles is moving in the correct direction to contribute to this flux I think it is saying $<v_{x}>= \frac{1}{2}<v>$. $\endgroup$ May 3, 2017 at 5:48

The probability density function (PDF) for a single component (let's take $z$) is

$$f(v_z)=\sqrt{\frac{m}{2\pi k T}} \exp \left(-\frac{m v_z^2}{2kT}\right)$$

Case #1

What you want to calculate is

$$\langle |v_z| \rangle = \int_{-\infty}^{\infty} |v_z| \ f(v_z) \ d v_z = 2 \int_0^\infty v_z \ f(v_z) \ d v_z$$


$$\int_0^{\infty} u e^{-au^2} du = \frac 1 {2a} \tag{1}$$

you get

$$\langle | v_z| \rangle = 2 \cdot \sqrt{\frac{m}{2\pi k T}} \cdot \frac 1 2 \cdot \frac{2kT}{m} = 2 \sqrt{\frac{kT}{2 \pi m}} = \sqrt{\frac{2kT}{\pi m}} $$

Notice that since

$$\langle |v| \rangle = \sqrt{\frac{8kT}{\pi m}}$$

where $|v|= \sqrt{v_x^2+v_y^2+v_z^2}$, you have

$$\langle | v_z| \rangle = \frac 1 2 \langle |v| \rangle $$

Case #2

As you said, in this case the normalization factor must change, since we must require that

$$\int_0^\infty \tilde f(v_z) d v_z = 1$$

The correct PDF is in this case

$$\tilde f(v_z)= \frac 1 2 \sqrt{\frac{m}{2\pi k T}} \exp \left(-\frac{m v_z^2}{2kT}\right)$$

Using again (1), it is straight-forward to obtain from this distribution

$$\langle v_z \rangle_{\tilde f} = \frac 1 2 \sqrt{\frac {kT}{2 \pi m}} = \frac 1 4 \langle |v_z| \rangle = \frac 1 8 \langle |v| \rangle $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.