0
$\begingroup$

I have numerically computed phonon dispersion using small displacement method by solving dynamical matrix along high symmetry direction [100], [110] and [111] for f.c.c lattice. Now I want to calculate density of state i.e $D(\omega)$ and plot $D(\omega)$ vs. $\omega$ in a graph. I have managed to do that by generating random $q$ vectors then computed phonon frequencies using my code and constructed a histogram using bin_size = 150 (though not sure what should be the range of $q$ vectors and the normalization of count). There exist a direct method to compute density of state for 1D phonon dispersion but don't know any existing method for 3D lattice. Can anyone please enlighten me in this matter, any reference or detail explanation will be very helpful.

Thanks

$\endgroup$
0

1 Answer 1

3
$\begingroup$

I will present the result for the one-dimensional case only. It is readily generalized to your three-dimensional situation. Knowing the phonon dispersion $\omega(q)$, the DOS can be defined as

$$D(\omega) = \frac{1}{2\pi}\int\mathrm{d}q\,\delta(\omega-\omega(q)).$$

Using the formal identity

$$\delta(g(x)) = \sum_{i}\frac{\delta(x-x_i)}{|g'(x_i)|}$$

where $x_i$ are the zeros of $g$ and $g'$ is its derivative, one can write:

$$D(\omega) = \frac{1}{2\pi}\int\mathrm dq \frac{1}{\left|\frac{\mathrm d\omega}{\mathrm dq}\right|}\sum_{\substack{i\\\omega(q_i)=\omega}}\delta(q-q_i) = \frac{1}{2\pi}\sum_{\substack{i\\\omega(q_i)=\omega}}\left|\frac{\mathrm d\omega}{\mathrm dq}(q_i)\right|^{-1}$$

The sum is over all the points where the dispersion reaches the given frequency $\omega$. The flater the dispersion at those points, the more states contribute!

$\endgroup$
2
  • $\begingroup$ Thanks. Its useful when we know the explicit form of function $\omega(q), but in case where we numerically compute phonon dispersion, how can I use this relation to compute density of state? If you can atleast show me any algorithm/pseudocode to compute Dos using this relation. $\endgroup$ Commented Aug 17, 2015 at 20:22
  • $\begingroup$ The above formula is suitable for numerical implementation. If you know the dispersion on a grid of $q$-points, you can approximate the derivative by a finite difference quotient. $\endgroup$ Commented Aug 18, 2015 at 8:07

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.