Take the 2-minute tour ×
Physics Stack Exchange is a question and answer site for active researchers, academics and students of physics. It's 100% free, no registration required.

In the textbook Understanding Molecular Simulation by Frenkel and Smit (Second Edition), the authors represent a function $f(\textbf{r})$ (which depends on the coordinates of a periodic system) as a Fourier series. I quote from page 295 of the text:

Let us consider a periodic system with a cubic box of length $L$ and volume $V$. Any function $f(\textbf{r})$ that depends on the coordinates of our system can be represented by a Fourier series:

$$f(\textbf{r}) = \frac{1}{V} \sum_{\boldsymbol{\ell} = -\infty}^{\infty} \tilde{f}(\textbf{k}) e^{i \textbf{k} \cdot \textbf{r}} \; \; \; \; \textbf{(12.1.6)}$$

where $\textbf{k} = \frac{2\pi}{L}\boldsymbol{\ell}$ with $\boldsymbol{\ell} = (\ell_x, \ell_y, \ell_z)$ are the lattice vectors in Fourier space. The Fourier coefficients $\tilde{f}(\textbf{k})$ are calculated using

$$\tilde{f}(\textbf{k}) = \int_V d\textbf{r} \; f(\textbf{r}) e^{-i\textbf{k} \cdot \textbf{r}} \; \; \; \; \textbf{(12.1.7)}$$

Now, the authors use equation (12.1.6) to write the electric potential $\phi(\textbf{r})$ in Fourier space:

$$\phi(\textbf{r}) = \frac{1}{V} \sum_{\textbf{k}} \tilde{\phi}(\textbf{k}) e^{i\textbf{k} \cdot \textbf{r}}$$

The authors write:

In Fourier space, Poisson's equation has a much simpler form. We can write for the Poisson equation:

$$-\nabla^2 \phi(\textbf{r}) = -\nabla^2 \left( \frac{1}{V} \sum_{\textbf{k}} \tilde{\phi}(\textbf{k}) e^{i\textbf{k} \cdot \textbf{r}} \right) = \frac{1}{V} \sum_{\textbf{k}} k^2 \tilde{\phi}(\textbf{k}) e^{i\textbf{k} \cdot \textbf{r}} \; \; \; \; \textbf{(12.1.8)}$$

My question is, why is the $\frac{1}{V}$ factor present in equations (12.1.6) and (12.1.8)? What is the significance of the $\frac{1}{V}$ factor in $\phi(\textbf{r}) = \frac{1}{V} \sum_{\textbf{k}} \tilde{\phi}(\textbf{k}) e^{i\textbf{k} \cdot \textbf{r}}$?

In contrast, the article on Wikipedia does not include this prefactor. I realize that that article is dealing with the general case, whereas here we are considering a system with a cubic box of volume $V$. But shouldn't the units of $\phi(\textbf{r})$ be the same as those of $\tilde{\phi}(\textbf{k})$? The $\frac{1}{V}$ seems to preclude $\phi(\textbf{r})$ and $\tilde{\phi}(\textbf{k})$ having the same units.

Do you have any advice? Thanks.

share|improve this question
add comment

2 Answers

up vote 2 down vote accepted

I) Let us just consider $1$ dimension for simplicity. (The generalization to higher dimensions is straightforward). Then the volume factor $V$ is just a length factor $L$.

II) The standard Fourier series formulas can be derived from $(12.1.7)$ and $(12.1.6)$ by taken the length $L$ to be $L=2\pi$. Then $(12.1.7)$ and $(12.1.6)$ become the standard Fourier series formulas

$$\tag{12.1.7'} c_{n} ~=~ \frac{1}{2\pi}\int_{-\pi}^{\pi} \! dx~ f(x) e^{-in x}, $$ $$\tag{12.1.6'} f(x)~=~\sum_{n\in\mathbb{Z}} c_n~e^{in x} ~=~f(x+2\pi), $$

via the identifications

$$\ell~=~ n~\in~\mathbb{Z}, \qquad \tilde{f}(\ell) ~=~2\pi c_{n} . $$

III) Going back to $3$ dimensions, the $1/V$ normalization in $(12.1.6)$ is important. Of course, in another convention, it could be put in $(12.1.7)$ instead, or alternatively, symmetrically as $1/\sqrt{V}$ in both formulas $(12.1.6)$ and $(12.1.7)$.

share|improve this answer
Thanks. But, given the $1/V$ normalization, what are the units of $\tilde{\phi}(\textbf{k})$ with respect to those of $\phi(\textbf{r})$ in the equation $$\phi(\textbf{r}) = \frac{1}{V} \sum_{\textbf{k}} \tilde{\phi}(\textbf{k}) e^{i\textbf{k}\cdot\textbf{r}}$$? –  Andrew Aug 9 '12 at 20:42
$[\tilde{f}]= [V] [f]= [L]^3 [f]$, cf. eq. $(12.1.7)$, or eq. $(12.1.6)$. –  Qmechanic Aug 9 '12 at 20:52
add comment

The usual Fourier series formula states that a function $g(x)$ with period $2\pi$ can be expressed as

$$ g(x) = \sum_{k\in {\mathbb Z}} c_k \, e^{ikx}$$

where $c_n$ (the coefficients of the Fourier series) are given by

$$c_k = \frac{1}{2 \pi} \int_{0}^{2 \pi} g(x) \, e^{-i x \, k}dx$$

If we have $f(x)$ with period $L$, we can relate with the above by $f(x) = g(2 \pi x/L)$, so we get the equivalent formulas:

$$ f(x) = \sum_{k\in {\mathbb Z}} c_k \, e^{i k x}$$

$$c_k = \frac{1}{L} \int_{0}^{L} f(x) \, e^{-i x \, k}dx$$

One can, alternatively move the normalization factor $1/L$ to the first equation. It's just a matter a convention.

The extension to 3 dimensions is straightforward.

share|improve this answer
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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