I am trying to derive a version of the Boltzmann distribution from statistical mechanics:

$$n(\mathbf{x}) = n_{0}e^{-q\phi(\mathbf{x})/kT}.$$

This expression is commonly used in the derivation of the Debye length in plasma physics (see Wikipedia).

I have found this article helpful. I have also looked at the discussion around equations (7.29)-(7.31) in Thermal Physics by Daniel Schroeder. However, the results in these sources don't exactly match the equation as given above. Both sources give:

$$P_i = \frac{\left< N_i \right>}{N} = \frac{1}{Z}e^{-\varepsilon_i/kT},$$

where $\left< N_i \right>$ is the average number of particles in the $i$th state. Obviously, this is close to the expression above.

I have tried the following with energies that are not discrete. I assume a canonical ensemble and a single particle of charge $q$ in a potential $\phi(\mathbf{x})$. The probability density is

$$P(\mathbf{x}, \mathbf{p}) = \frac{\exp\left[-q \phi(\mathbf{x})/kT\right] \exp\left[-p^2 / 2mkT\right]}{\int d\mathbf{x} \, d\mathbf{p} \, \exp\left[-q \phi(\mathbf{x})/kT\right] \exp\left[-p^2 / 2mkT\right]}$$

Integrating over momentum space gives the probability density over configuration space:

$$\int d\mathbf{p} \, P(\mathbf{x}, \mathbf{p}) = P(\mathbf{x}) = \frac{\exp\left[-q \phi(\mathbf{x})/kT\right]}{\int d\mathbf{x} \, \exp\left[-q \phi(\mathbf{x})/kT\right]}$$

I interpret this as the probability per unit volume that particle q will be located at $\mathbf{x}$. If I multiply now by the number of particles, $N$, I have the average number of particles per unit volume at $\mathbf{x}$, which I interpret as $n(\mathbf{x})$:

$$n(\mathbf{x}) = N \frac{\exp\left[-q \phi(\mathbf{x})/kT\right]}{\int d\mathbf{x} \, \exp\left[-q \phi(\mathbf{x})/kT\right]. }$$

Comparing to the equation that I gave above, it seems that I am close. What am I missing?


1 Answer 1


I believe that the steps in the posted question are (more or less) correct. The final connection between the first and last equations can be seen by integrating the first equation:

$$\int d\mathbf{x} \, n(\mathbf{x}) = \int d\mathbf{x} \, n_0 e^{-q\Phi(\mathbf{x})/kT}$$ $$N = n_0 \int d\mathbf{x} \, e^{-q\Phi(\mathbf{x})/kT}$$

Solving for $n_0$ gives

$$n_0 = \frac{N}{\int d\mathrm{x} \, \exp(-q\Phi(\mathbf{x})/kT)}$$

Comparing this to the last equation in the original question, we see that the last equation is in fact identical to the first equation.

As a side note, for $q\Phi(\mathbf{x}) \ll kT$ the density becomes $$n(\mathbf{x}) = n_0$$ If the source of $\Phi$ is at the origin, then this approximation is valid for $\mathbf{x} \to \infty$ For this reason, F.F. Chen (Introduction to Plasma Physics and Controlled Fusion) uses $n_\infty$ for $n_0$.

Also, Chapter 4 of Statistical Mechanics by McQuarrie gives a very extensive discussion of how to transition from the probability that a system will be in a given state to the probability that a particle will be in a given state. This helps justify using a single particle in the original question. The assumptions are that the particles do not interact, temperature is high, and density is low. The first assumption of non-interacting particles is perhaps not so great for charged particles in a plasma. However, if the high temperature and low density conditions are met, then the assumption of non-interaction becomes better.

I would like to thank one of my colleagues for helping me with this. He could have posted the answer, but he doesn't have an account. Pity :)


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.