Wave propagation is characterized by the wavenumber $k$ and the angular frequency $\omega$. Sometimes (like in this answer) the relation $\omega (k)$ is preferred; sometimes instead $k ( \omega)$ is used.

When expanding $k(\omega)$ around a point $k = k_0$:

$$\omega (k) \simeq \omega(k_0) + (k - k_0) \left. \frac{d\omega(k)}{d k} \right|_{k = k_0} + \frac{1}{2} (k - k_0)^2 \left. \frac{d^2 \omega(k)}{dk^2} \right|_{k = k_0}$$

or vice-versa:

$$k (\omega) \simeq k(\omega_0) + (\omega - \omega_0) \left. \frac{dk(\omega)}{d \omega} \right|_{\omega = \omega_0} + \frac{1}{2} (\omega - \omega_0)^2 \left. \frac{d^2 k(\omega)}{d\omega^2} \right|_{\omega = \omega_0}$$

$\left. \frac{d\omega(k)}{d k} \right|_{k = k_0} = v_g$ is the group velocity in the first expansion; instead $\left. \frac{dk (\omega)}{d \omega} \right|_{\omega = \omega_0} = 1/v_g$ is its reciprocal.

But what does happen if $v_g \to 0$? Will the second expansion have an infinite first-order coefficient? The same question is for the second derivative quantities. The two expansions should be interchangeable, anyway they seem to be different.


Let us write the dispersion relation in a mathematically clearer version :

$$ \omega = f(k)$$ where we note $f$ the dispersion function.

Say you want to compute the group velocity at $k_0$. It is defined as:

$$ v_g(k_0) \equiv \frac{df}{dk}(k_0) = \left.\frac{d\omega(k)}{dk}\right|_{k = k_0} \text{ (in your notation)}$$

Up to this point this is a definition and if $v_g(k_0) \rightarrow0$ it means that the derivative of the dispersion function $f$ at $k_0$ is $0$, which is completely possible.

Now let us look at the inverse dispersion function $g$ defined by:


If $f$ is invertible on a small interval around $k_0$ (say $[k_0-\epsilon,k_0+\epsilon]$ with $\epsilon \in \mathbb{R}$ ), then in this interval we can write $g=f^{-1}$ and there is a unique $\omega_0$ in the interval that verifies $k_0 = g(\omega_0)$ and $\omega_0 = f(k_0)$. In the case that $\frac{df}{dk}(k_0) \neq 0$, you have the following formula for the derivative of $g$ :

$$\left.\frac{dk(\omega)}{d\omega}\right|_{\omega = \omega_0} = \frac{dg}{d\omega}(\omega_0) = \frac{1}{\frac{df}{dk}(g(\omega_0))} = \frac{1}{\frac{df}{dk}(k_0)} = \frac{1}{v_g(k_0)}$$

Note that this formula is only valid when $v_g(k_0) = \frac{df}{dk}(k_0) \neq 0$. This means that when you say that the two terms should be interchangeable, this is not mathematically true. It is only the case if the dispersion function $f$ is invertible around $k_0$ and its derivative at $k_0$ non zero. This solves mathematically the paradox that you are evoking.

To sum it up, it means that you have to chose a definition for the group velocity, and the most common one to my knowledge is $ v_g \equiv\left.\frac{d\omega(k)}{dk}\right|_{k = k_0}$. If it goes to zero at a certain value $k_0$ it means that the dispersion function has a horizontal tangent at this point, and that the reciprocal dispersion function has an infinite slope at a given value $\omega_0$. This doesn't imply anything on definition of the group velocity, since it is a definition !

I hope my answer was clear enough.



Imagine a plot $\omega(k)$ where, for some $k_0$, $v_g(k_0) = \left. \frac{\mathrm d\omega}{\mathrm dk} \right|_{k_0} = 0$. Then the graph will have a horizontal tangent at $k=k_0$, like in this example:

enter image description here

The graph of $k(\omega)$ is just the mirror image, obviously it will have a vertical tangent at the respective point:

enter image description here

So yes, the first derivative diverges and the function $k(\omega)$ can not be expanded around this point.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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